Method and system for locating objects in an image

• US 6,072,893 A
• Filed: 01/06/1999
• Issued: 06/06/2000
• Est. Priority Date: 08/29/1997
• Status: Expired due to Term

• Current Assignee: Monument Peak Ventures, LLC (Dominion Harbor Enterprises, LLC)
• Original Assignee: Eastman Kodak Company
• Inventors: Luo, Jiebo, Gray, Robert T.
• Primary Examiner(s): Au, Amelia
• Assistant Examiner(s): Wu, Jingge
• Application Number:
  US09/226,325
• Time in Patent Office: 517 Days
• Field of Search: 382/117, 382/118, 382/291
• US Class Current: 382/117
• CPC Class Codes: G06T 2207/30216 Redeye defect
• CPC Class Codes: G06T 7/74 involving reference images or patches
• CPC Class Codes: G06V 40/19 Sensors therefor

Abstract

A computer program product for locating first and second objects each having substantially the same physical characteristics, and the ratio of the distance between the first and second objects and the size of each object is substantially invariant, the computer program product comprises: a computer readable storage medium having a computer program stored thereon for performing the steps of: displaying a digital image on a graphical user interface; inputting to the interface an approximate position of each object for forming initial input positions; determining an estimate size of each object based on the distance obtained from the initial input positions of the first and second objects; forming a first searching window for the first object; the center of the first window is determined by the input position of the first object and the size of the first window is determined by the estimate size of the first object; reiteratively positioning a template on the first search window for determining a location that gives a best match of the first object relative to the template; forming a second searching window for the second object; the center of the second window is determined by the input position of the second object and the size of the second window is determined by the estimate size of the second object; and reiteratively positioning a template on the second search window for determining a location that gives a best match of the second object relative to the template.

Claims

1. A method for locating first and second objects each having substantially the same physical characteristics, and the ratio of the distance between the first and second objects and the size of each object is substantially invariant, the method comprising the steps of:

(a) displaying a digital image on a graphical user interface;

(b) inputting to the interface an approximate position of each object for forming initial input positions;

(c) determining an estimate size of each object based on the distance obtained from the initial input positions of the first and second objects;

(d) forming a first searching window for the first object;

the center of the first window is determined by the input position of the first object and the size of the first window is determined by the estimate size of the first object;

(e) reiteratively positioning a template on the first search window for determining a location that gives a desired match of the first object relative to the template;

(f) forming a second searching window for the second object;

the center of the second window is determined by the input position of the second object and the size of the second window is determined by the estimate size of the second object; and

(g) reiteratively positioning a template on the second search window for determining a location that gives a desired match of the second object relative to the template.

2. The method as in claim 1, further comprising the step of (h) determining a plurality of locations in each search window that give a desired matching response to the template.

3. The method as in claim 2 further comprising the step of finding the best pair of locations from all possible combinations of identified locations of both the first and second objects.

4. The method as in claim 3, wherein the step of finding best pair of locations includes computing a plurality of figures of merit and determining a pair of locations that give a best combined figure of merit.

5. The method as in claim 4, wherein computing a plurality of figures of merit includes computing individually or in combination an orientation, proximity, combined correlation score, proportion, profile, symmetry or evidence of mouth.

6. The method as in claim 5, wherein the first and second objects are first and second human eyes and wherein computing an orientation includes measuring the difference between the orientation of the line connecting the first and second eyes, and the average orientation of the first and second eyes.

7. The method as in claim 5, wherein computing a proximity includes measuring proximity of the first and second objects to the initial input positions, respectively.

8. The method as in claim 5, wherein computing a combined correlation score includes summing up the individual correlation scores obtained from the matching obtained in step (e1).

9. The method as in claim 5, wherein the first and second objects are first and second human eyes and wherein computing a proportion includes computing a distance between the first and second eyes for satisfying an anthropological prior model.

10. The method as in claim 5, wherein the first and second objects are first and second human eyes and wherein computing the profile includes pre-determining a model of the eye-to-eye profile and determining an actual profile from the image and computing a goodness-of-fit between the actual profile and the model profile.

11. The method as in claim 5, wherein the first and second objects are first and second human eyes and wherein computing the symmetry includes computing a symmetry between first and second halves of a face window determined by the eye locations.

12. The method as in claim 5, wherein the first and second objects are first and second human eyes and wherein computing the evidence of mouth includes computing strength and orientation of edges within a mouth window determined by the eye locations.

13. The method as in claim 2 further comprising the step of normalizing the template for maximizing robustness of matching.

14. The method as in claim 2 further comprising the step of creating a image block of the same size and orientation as the template.

15. The method as in claim 14 further comprising the step of normalizing the image block for maximizing robustness of matching.

16. The method as in claim 14 further comprising the step of extracting individual sub-zones in both the template and the image block, and performing cross-correlation for each corresponding sub-zone.

17. The method as in claim 14 further comprising the step of computing the overall cross-correlation score for the image block if the cross-correlation score for each sub-zone exceeds a threshold.

18. The method as in claim 2 further comprising the step of identifying the locations that give locally maximum cross-correlation score.

19. A method for locating first and second objects each having substantially the same physical characteristics, and the ratio of the distance between the first and second objects and the size of each object is substantially invariant, the method comprising the steps of:

(a) displaying a digital image on a graphical user interface;

(b) inputting to the interface an approximate position of each object for forming initial input positions;

(c) determining an estimate orientation of each object based on an angle derived from the initial input positions of the first and second objects;

(d) forming a first searching window for the first object;

the center of the first window is determined by the initial input position of the first object and the size of the first window is determined by the estimate size of the first object;

(e) reiteratively positioning a template on the first search window for determining a location that gives a desired match of the first object relative to the template; and

(f) forming a second searching window for the first object;

the center of the second window is determined by the input position of the first object and the size of the second window is determined by the estimate size of the first object;

(g) reiteratively positioning a template on the second search window for determining a location that gives a desired match of the first object relative to the template.

20. The method as in claim 19 further comprising determining a plurality of locations that give a desired matching response to the template.

21. The method as in claim 19 further comprising the step of finding the best pair of locations from all possible combinations of identified locations of both the first and second objects.

22. The method as in claim 21, wherein the step of finding best pair of locations includes computing a plurality of figures of merit and determining a pair of locations that give a best combined figure of merit.

23. The method as in claim 22, wherein computing a plurality of figures of merit includes computing individually or in combination an orientation, proximity, combined correlation score, proportion, profile, symmetry or evidence of mouth.

24. The method as in claim 23, wherein the first and second objects are first and second human eyes and wherein computing an orientation includes measuring the difference between the orientation of the line connecting the first and second eyes, and the average orientation of the first and second eyes.

25. The method as in claim 23, wherein computing a proximity includes measuring proximity of the first and second objects to the initial input positions, respectively.

26. The method as in claim 23, wherein computing a combined correlation score includes summing up the individual correlation scores obtained from the matching obtained in step (e1).

27. The method as in claim 23, wherein the first and second objects are first and second human eyes and wherein computing a proportion includes computing a distance between the first and second eyes for satisfying an anthropological prior model.

28. The method as in claim 23, wherein the first and second objects are first and second human eyes and wherein computing the profile includes pre-determining a model of the eye-to-eye profile and determining an actual profile from the image and computing a goodness-of-fit between the actual profile and the model profile.

29. The method as in claim 23, wherein the first and second objects are first and second human eyes and wherein computing the symmetry includes computing a symmetry between first and second halves of a face window determined by the eye locations.

30. The method as in claim 23, wherein the first and second objects are first and second human eyes and wherein computing the evidence of mouth includes computing strength and orientation of edges within a mouth window determined by the eye locations.

31. The method as in claim 20 further comprising the step of normalizing the template for maximizing robustness of matching.

32. The method as in claim 20 further comprising the step of creating a image block of the same size and orientation as the template.

33. The method as in claim 32 further comprising the step of normalizing the image block for maximizing robustness of matching.

34. The method as in claim 32 further comprising the step of extracting individual sub-zones in both the template and the image block, and performing cross-correlation for each corresponding sub-zone.

35. The method as in claim 32 further comprising the step of computing the overall cross-correlation score for the image block if the cross-correlation score for each sub-zone exceeds a threshold.

36. The method as in claim 20 further comprising the step of identifying the locations that give locally maximum cross-correlation score.

37. A method for locating first and second objects each having substantially the same physical characteristics, and the ratio of the distance between the first and second objects and the size of each object is substantially invariant, the method comprising the steps of:

(a) displaying a digital image on a graphical user interface;

(b) inputting to the interface an approximate position of each object for forming initial input positions;

(c) determining an estimate orientation of each object based on an angle derived from the initial input positions of the first and second objects;

(d) determining an estimate size of each object based on the distance obtained from the initial input positions of the first and second objects;

(e) forming a searching window for the first object;

the center of the window is determined by the input position of the first object and the size of the window is determined by the estimate size of the first object;

(f) reiteratively positioning a template on the search window for determining a location that gives a desired match of the first object relative to the template; and

(g) repeating steps (e) and (f) for the second object.

38. The method as in claim 37, further comprising the step of (h) determining a plurality of locations in each search window that give a desired matching response to the template.

39. The method as in claim 38 further comprising the step of finding the best pair of locations from all possible combinations of identified locations of both the first and second objects.

40. The method as in claim 39, wherein the step of finding best pair of locations includes computing a plurality of figures of merit and determining a pair of locations that give a best combined figure of merit.

41. The method as in claim 40, wherein computing a plurality of figures of merit includes computing individually or in combination an orientation, proximity, combined correlation score, proportion, profile, symmetry or evidence of mouth.

42. The method as in claim 41, wherein the first and second objects are first and second human eyes and wherein computing an orientation includes measuring the difference between the orientation of the line connecting the first and second eyes, and the average orientation of the first and second eyes.

43. The method as in claim 41, wherein computing a proximity includes measuring proximity of the first and second objects to the initial input positions, respectively.

44. The method as in claim 41, wherein computing a combined correlation score includes summing up the individual correlation scores obtained from the matching obtained in step (e1).

45. The method as in claim 41, wherein the first and second objects are first and second human eyes and wherein computing a proportion includes computing a distance between the first and second eyes for satisfying an anthropological prior model.

46. The method as in claim 41, wherein the first and second objects are first and second human eyes and wherein computing the profile includes pre-determining a model of the eye-to-eye profile and determining an actual profile from the image and computing a goodness-of-fit between the actual profile and the model profile.

47. The method as in claim 41, wherein the first and second objects are first and second human eyes and wherein computing the symmetry includes computing a symmetry between first and second halves of a face window determined by the eye locations.

48. The method as in claim 41, wherein the first and second objects are first and second human eyes and wherein computing the evidence of mouth includes computing strength and orientation of edges within a mouth window determined by the eye locations.

49. The method as in claim 38 further comprising the step of normalizing the template for maximizing robustness of matching.

50. The method as in claim 38 further comprising the step of creating a image block of the same size and orientation as the template.

51. The method as in claim 50 further comprising the step of normalizing the image block for maximizing robustness of matching.

52. The method as in claim 50 further comprising the step of extracting individual sub-zones in both the template and the image block, and performing cross-correlation for each corresponding sub-zone.

53. The method as in claim 50 further comprising the step of computing the overall cross-correlation score for the image block if the cross-correlation score for each sub-zone exceeds a threshold.

54. The method as in claim 38 further comprising the step of identifying the locations that give locally maximum cross-correlation score.

55. A method for locating first and second objects in an image each object having substantially the same physical characteristics, and the ratio of the distance between the first and second objects and the size of each object is substantially invariant, the method comprising the steps of:

(a) displaying a digital image on a graphical user interface;

(b) estimating an approximate position of each object for forming initial input positions based on a format of the image;

(c) determining an estimate size of each object based on the distance obtained from the initial input positions of the first and second objects;

(d) forming a first searching window for the first object;

the center of the first window is determined by the input position of the first object and the size of the first window is determined by the estimate size of the first object;

(e) reiteratively positioning a template on the first search window for determining a location that gives a desired match of the first object relative to the template;

(f) forming a second searching window for the second object;

the center of the second window is determined by the input position of the second object and the size of the second window is determined by the estimate size of the second object; and

(g) reiteratively positioning a template on the second search window for determining a location that gives a desired match of the second object relative to the template.

56. The method as in claim 55, further comprising the step of (h) determining a plurality of locations in each search window that give a desired matching response to the template.

57. The method as in claim 56 further comprising the step of finding the best pair of locations from all possible combinations of identified locations of both the first and second objects.

58. The method as in claim 57, wherein the step of finding best pair of locations includes computing a plurality of figures of merit and determining a pair of locations that give a best combined figure of merit.

59. The method as in claim 58, wherein computing a plurality of figures of merit includes computing individually or in combination an orientation, proximity, combined correlation score, proportion, profile, symmetry or evidence of mouth.

60. The method as in claim 59, wherein the first and second objects are first and second human eyes and wherein computing an orientation includes measuring the difference between the orientation of the line connecting the first and second eyes, and the average orientation of the first and second eyes.

61. The method as in claim 59, wherein computing a proximity includes measuring proximity of the first and second objects to the initial input positions, respectively.

62. The method as in claim 59, wherein computing a combined correlation score includes summing up the individual correlation scores obtained from the matching obtained in step (e).

63. The method as in claim 59, wherein the first and second objects are first and second human eyes and wherein computing a proportion includes computing a distance between the first and second eyes for satisfying an anthropological prior model.

64. The method as in claim 59, wherein the first and second objects are first and second human eyes and wherein computing the profile includes pre-determining a model of the eye-to-eye profile and determining an actual profile from the image and computing a goodness-of-fit between the actual profile and the model profile.

65. The method as in claim 59, wherein the first and second objects are first and second human eyes and wherein computing the symmetry includes computing a symmetry between first and second halves of a face window determined by the eye locations.

66. The method as in claim 59, wherein the first and second objects are first and second human eyes and wherein computing the evidence of mouth includes computing strength and orientation of edges within a mouth window determined by the eye locations.

67. The method as in claim 56 further comprising the step of normalizing the template for maximizing robustness of matching.

68. The method as in claim 56 further comprising the step of creating a image block of the same size and orientation as the template.

69. The method as in claim 68 further comprising the step of normalizing the image block for maximizing robustness of matching.

70. The method as in claim 68 further comprising the step of extracting individual sub-zones in both the template and the image block, and performing cross-correlation for each corresponding sub-zone.

71. The method as in claim 68 further comprising the step of computing the overall cross-correlation score for the image block if the cross-correlation score for each sub-zone exceeds a threshold.

72. The method as in claim 56 further comprising the step of identifying the locations that give locally maximum cross-correlation score.

73. A method for locating first and second objects each having substantially the same physical characteristics, and the ratio of the distance between the first and second objects and the size of each object is substantially invariant, the method comprising the steps of:

(a) displaying a digital image on a graphical user interface;

(b) estimating an approximate position of each object for forming initial input positions;

(c) determining an estimate orientation of each object based on an angle derived from the initial input positions of the first and second objects;

(d) forming a first searching window for the first object;

the center of the first window is determined by the initial input position of the first object and the size of the first window is determined by the estimate size of the first object;

(e) reiteratively positioning a template on the first search window for determining a location that gives a desired match of the first object relative to the template; and

(f) forming a second searching window for the first object;

the center of the second window is determined by the input position of the first object and the size of the second window is determined by the estimate size of the first object;

(g) reiteratively positioning a template on the second search window for determining a location that gives a desired match of the first object relative to the template.

74. The method as in claim 73 further comprising determining a plurality of locations that give a desired matching response to the template.

75. The method as in claim 73 further comprising the step of finding the best pair of locations from all possible combinations of identified locations of both the first and second objects.

76. The method as in claim 75, wherein the step of finding best pair of locations includes computing a plurality of figures of merit and determining a pair of locations that give a best combined figure of merit.

77. The method as in claim 76, wherein computing a plurality of figures of merit includes computing individually or in combination an orientation, proximity, combined correlation score, proportion, profile, symmetry or evidence of mouth.

78. The method as in claim 77, wherein the first and second objects are first and second human eyes and wherein computing an orientation includes measuring the difference between the orientation of the line connecting the first and second eyes, and the average orientation of the first and second eyes.

79. The method as in claim 77, wherein computing a proximity includes measuring proximity of the first and second objects to the initial input positions, respectively.

80. The method as in claim 77, wherein computing a combined correlation score includes summing up the individual correlation scores obtained from the matching obtained in step (e1).

81. The method as in claim 77, wherein the first and second objects are first and second human eyes and wherein computing a proportion includes computing a distance between the first and second eyes for satisfying an anthropological prior model.

82. The method as in claim 77, wherein the first and second objects are first and second human eyes and wherein computing the profile includes pre-determining a model of the eye-to-eye profile and determining an actual profile from the image and computing a goodness-of-fit between the actual profile and the model profile.

83. The method as in claim 77, wherein the first and second objects are first and second human eyes and wherein computing the symmetry includes computing a symmetry between first and second halves of a face window determined by the eye locations.

84. The method as in claim 77, wherein the first and second objects are first and second human eyes and wherein computing the evidence of mouth includes computing strength and orientation of edges within a mouth window determined by the eye locations.

85. The method as in claim 74 further comprising the step of normalizing the template for maximizing robustness of matching.

86. The method as in claim 74 further comprising the step of creating a image block of the same size and orientation as the template.

87. The method as in claim 86 further comprising the step of normalizing the image block for maximizing robustness of matching.

88. The method as in claim 87 further comprising the step of extracting individual sub-zones in both the template and the image block, and performing cross-correlation for each corresponding sub-zone.

89. The method as in claim 87 further comprising the step of computing the overall cross-correlation score for the image block if the cross-correlation score for each sub-zone exceeds a threshold.

90. The method as in claim 75 further comprising the step of identifying the locations that give locally maximum cross-correlation score.

91. A method for locating first and second objects in an image each object having substantially the same physical characteristics, and the ratio of the distance between the first and second objects and the size of each object is substantially invariant, the method comprising the steps of:

(a) displaying a digital image on a graphical user interface;

(b) estimating an approximate position of each object for forming initial input positions based on a format of the image;

(c) determining an estimate orientation of each object based on an angle derived from the initial input positions of the first and second objects;

(d) determining an estimate size of each object based on the distance obtained from the initial input positions of the first and second objects;

(e) forming a searching window for the first object;

the center of the window is determined by the input position of the first object and the size of the window is determined by the estimate size of the first object;

(f) reiteratively positioning a template on the search window for determining a location that gives a desired match of the first object relative to the template; and

(g) repeating steps (e) and (f) for the second object.

92. The method as in claim 91, further comprising the step of (h) determining a plurality of locations in each search window that give a desired matching response to the template.

93. The method as in claim 92 further comprising the step of finding the best pair of locations from all possible combinations of identified locations of both the first and second objects.

94. The method as in claim 93, wherein the step of finding best pair of locations includes computing a plurality of figures of merit and determining a pair of locations that give a best combined figure of merit.

95. The method as in claim 93, wherein computing a plurality of figures of merit includes computing individually or in combination an orientation, proximity, combined correlation score, proportion, profile, symmetry or evidence of mouth.

96. The method as in claim 95, wherein the first and second objects are first and second human eyes and wherein computing an orientation includes measuring the difference between the orientation of the line connecting the first and second eyes, and the average orientation of the first and second eyes.

97. The method as in claim 95, wherein computing a proximity includes measuring proximity of the first and second objects to the initial input positions, respectively.

98. The method as in claim 95, wherein computing a combined correlation score includes summing up the individual correlation scores obtained from the matching obtained in step (e1).

99. The method as in claim 95, wherein the first and second objects are first and second human eyes and wherein computing a proportion includes computing a distance between the first and second eyes for satisfying an anthropological prior model.

100. The method as in claim 95, wherein the first and second objects are first and second human eyes and wherein computing the profile includes pre-determining a model of the eye-to-eye profile and determining an actual profile from the image and computing a goodness-of-fit between the actual profile and the model profile.

101. The method as in claim 95, wherein the first and second objects are first and second human eyes and wherein computing the symmetry includes computing a symmetry between first and second halves of a face window determined by the eye locations.

102. The method as in claim 95, wherein the first and second objects are first and second human eyes and wherein computing the evidence of mouth includes computing strength and orientation of edges within a mouth window determined by the eye locations.

103. The method as in claim 92 further comprising the step of normalizing the template for maximizing robustness of matching.

104. The method as in claim 92 further comprising the step of creating a image block of the same size and orientation as the template.

105. The method as in claim 104 further comprising the step of normalizing the image block for maximizing robustness of matching.

106. The method as in claim 104 further comprising the step of extracting individual sub-zones in both the template and the image block, and performing cross-correlation for each corresponding sub-zone.

107. The method as in claim 104 further comprising the step of computing the overall cross-correlation score for the image block if the cross-correlation score for each sub-zone exceeds a threshold.

108. The method as in claim 92 further comprising the step of identifying the locations that give locally maximum cross-correlation score.

110. The system as in claim 108, further comprising (h) means for determining a plurality of locations in each search window that give a desired matching response to the template.

111. The system as in claim 110 further comprising means for finding the best pair of locations from all possible combinations of identified locations of both the first and second objects .

112. The system as in claim 111, wherein the means for finding best pair of locations includes means for computing a plurality of figures of merit and determining a pair of locations that give a best combined figure of merit.

113. The system as in claim 112, wherein the means for computing a plurality of figures of merit includes means for computing individually or in combination an orientation, proximity, combined correlation score, proportion, profile, symmetry or evidence of mouth.

114. The system as in claim 113, wherein the first and second objects are first and second human eyes and wherein computing an orientation includes measuring the difference between the orientation of the line connecting the first and second eyes, and the average orientation of the first and second eyes.

115. The system as in claim 113, wherein computing a proximity includes measuring proximity of the first and second objects to the initial input positions, respectively.

116. The system as in claim 113, wherein the means for computing a combined correlation score includes means for summing up the individual correlation scores obtained from the matching obtained in (e).

117. The system as in claim 113, wherein the first and second objects are first and second human eyes and wherein means for computing a proportion includes means for computing a distance between the first and second eyes for satisfying an anthropological prior model.

118. The system as in claim 113, wherein the first and second objects are first and second human eyes and wherein the means for computing the profile includes means for pre-determining a model of the eye-to-eye profile and means for determining an actual profile from the image and means for computing a goodness-of-fit between the actual profile and the model profile.

119. The system as in claim 113, wherein the first and second objects are first and second human eyes and wherein the means for computing the symmetry includes means for computing a symmetry between first and second halves of a face window determined by the eye locations.

120. The system as in claim 113, wherein the first and second objects are first and second human eyes and wherein the means for computing the evidence of mouth includes means for computing strength and orientation of edges within a mouth window determined by the eye locations.

121. The system as in claim 110 further comprising means for normalizing the template for maximizing robustness of matching.

122. The system as in claim 110 further comprising means for creating a image block of the same size and orientation as the template.

123. The system as in claim 122 further comprising means for normalizing the image block for maximizing robustness of matching.

124. The system as in claim 122 further comprising means for extracting individual sub-zones in both the template and the image block, and means for performing cross-correlation for each corresponding sub-zone.

125. The system as in claim 122 further comprising means for computing overall cross-correlation score for the image block if the cross-correlation score for each sub-zone exceeds a threshold.

126. The system as in claim 110 further comprising means for identifying the locations that give locally maximum cross-correlation score.

109. A system for locating first and second objects each having substantially the same physical characteristics, and the ratio of the distance between the first and second objects and the size of each object is substantially invariant, the system comprising:

(a) means for displaying a digital image on a graphical user interface;

(b) means for inputting to the interface an approximate position of each object for forming initial input positions;

(c) means for determining an estimate size of each object based on the distance obtained from the initial input positions of the first and second objects;

(d) means for forming a first searching window for the first object;

the center of the first window is determined by the input position of the first object and the size of the first window is determined by the estimate size of the first object;

(e) means for reiteratively positioning a template on the first search window for determining a location that gives a desired match of the first object relative to the template;

(f) means for forming a second searching window for the second object;

the center of the second window is determined by the input position of the second object and the size of the second window is determined by the estimate size of the second object; and

(g) means for reiteratively positioning a template on the second search window for determining a location that gives a desired match of the second object relative to the template.

127. A system for locating first and second objects each having substantially the same physical characteristics, and the ratio of the distance between the first and second objects and the size of each object is substantially invariant, the system comprising:

(a) means for displaying a digital image on a graphical user interface;

(b) means for inputting to the interface an approximate position of each object for forming initial input positions;

(c) means for determining an estimate orientation of each object based on an angle derived from the initial input positions of the first and second objects;

(d) means for forming a first searching window for the first object;

the center of the first window is determined by the initial input position of the first object and the size of the first window is determined by the estimate size of the first object;

(e) means for reiteratively positioning a template on the first search window for determining a location that gives a desired match of the first object relative to the template; and

(f) means for forming a second searching window for the first object;

the center of the second window is determined by the input position of the first object and the size of the second window is determined by the estimate size of the first object;

(g) means for reiteratively positioning a template on the second search window for determining a location that gives a desired match of the first object relative to the template.

128. The system as in claim 127 further comprising means for determining a plurality of locations that give a desired matching response to the template.

129. The system as in claim 127 further comprising means for finding the best pair of locations from all possible combinations of identified locations of both the first and second objects.

130. The system as in claim 129, wherein the means for finding best pair of locations includes means for computing a plurality of figures of merit and means for determining a pair of locations that give a best combined figure of merit.

131. The system as in claim 130, wherein means for computing a plurality of figures of merit includes means for computing individually or in combination an orientation, proximity, combined correlation score, proportion, profile, symmetry or evidence of mouth.

132. The system as in claim 131, wherein the first and second objects are first and second human eyes and wherein the means for computing an orientation includes means for measuring the difference between the orientation of the line connecting the first and second eyes, and means for the average orientation of the first and second eyes.

133. The system as in claim 131, wherein the means for computing a proximity includes means for measuring proximity of the first and second objects to the initial input positions, respectively.

134. The system as in claim 131, wherein the means for computing a combined correlation score includes means for summing up the individual correlation scores obtained from the matching obtained in (e).

135. The system as in claim 131, wherein the first and second objects are first and second human eyes and wherein the means for computing a proportion includes means for computing a distance between the first and second eyes for satisfying an anthropological prior model.

136. The system as in claim 131, wherein the first and second objects are first and second human eyes and wherein the means for computing the profile includes means for pre-determining a model of the eye-to-eye profile and means for determining an actual profile from the image and computing a goodness-of-fit between the actual profile and the model profile.

137. The system as in claim 131, wherein the first and second objects are first and second human eyes and wherein the means for computing the symmetry includes means for computing a symmetry between first and second halves of a face window determined by the eye locations.

138. The system as in claim 131, wherein the first and second objects are first and second human eyes and wherein the means for computing the evidence of mouth includes means for computing strength and orientation of edges within a mouth window determined by the eye locations.

139. The system as in claim 128 further comprising means for normalizing the template for maximizing robustness of matching.

140. The system as in claim 128 further comprising the step of creating a image block of the same size and orientation as the template.

141. The system as in claim 140 further comprising means for normalizing the image block for maximizing robustness of matching.

142. The system as in claim 140 further comprising means for extracting individual sub-zones in both the template and the image block, and means for performing cross-correlation for each corresponding sub-zone.

143. The system as in claim 140 further comprising means for computing the overall cross-correlation score for the image block if the cross-correlation score for each sub-zone exceeds a threshold.

144. The system as in claim 128 further comprising means for identifying the locations that give locally maximum cross-correlation score.

145. A system for locating first and second objects each having substantially the same physical characteristics, and the ratio of the distance between the first and second objects and the size of each object is substantially invariant, the method comprising the steps of:

(a) means for displaying a digital image on a graphical user interface;

(b) means for inputting to the interface an approximate position of each object for forming initial input positions;

(c) means for determining an estimate orientation of each object based on an angle derived from the initial input positions of the first and second objects;

(d) means for determining an estimate size of each object based on the distance obtained from the initial input positions of the first and second objects;

(e) means for forming a searching window for the first object;

the center of the window is determined by the input position of the first object and the size of the window is determined by the estimate size of the first object;

(f) means for reiteratively positioning a template on the search window for determining a location that gives a desired match of the first object relative to the template; and

(g) means for repeating steps (e) and (f) for the second object.

146. The system as in claim 145, further comprising the step of (h) determining a plurality of locations in each search window that give a desired matching response to the template.

147. The system as in claim 146 further comprising means for finding the best pair of locations from all possible combinations of identified locations of both the first and second objects.

148. The system as in claim 147, wherein the means for finding best pair of locations includes means for computing a plurality of figures of merit and means for determining a pair of locations that give a best combined figure of merit.

149. The system as in claim 148, wherein the means for computing a plurality of figures of merit includes means for computing individually or in combination an orientation, proximity, combined correlation score, proportion, profile, symmetry or evidence of mouth.

150. The system as in claim 149, wherein the first and second objects are first and second human eyes and wherein computing an orientation includes measuring the difference between the orientation of the line connecting the first and second eyes, and the average orientation of the first and second eyes.

151. The system as in claim 149, wherein the means for computing a proximity includes means for measuring proximity of the first and second objects to the initial input positions, respectively.

152. The system as in claim 149, wherein the means for computing a combined correlation score includes means for summing up the individual correlation scores obtained from the matching obtained in (e).

153. The system as in claim 149, wherein the first and second objects are first and second human eyes and wherein the means for computing a proportion includes means for computing a distance between the first and second eyes for satisfying an anthropological prior model.

154. The system as in claim 149, wherein the first and second objects are first and second human eyes and wherein the means for computing the profile includes means for predetermining a model of the eye-to-eye profile and means for determining an actual profile from the image and computing a goodness-of-fit between the actual profile and the model profile.

155. The system as in claim 149, wherein the first and second objects are first and second human eyes and wherein computing the symmetry includes computing a symmetry between first and second halves of a face window determined by the eye locations.

156. The system as in claim 149, wherein the first and second objects are first and second human eyes and wherein computing the evidence of mouth includes computing strength and orientation of edges within a mouth window determined by the eye locations.

157. The system as in claim 146 further comprising means for normalizing the template for maximizing robustness of matching.

158. The system as in claim 146 further comprising the step of creating a image block of the same size and orientation as the template.

159. The system as in claim 158 further comprising means for normalizing the image block for maximizing robustness of matching.

160. The system as in claim 158 further comprising means for extracting individual sub-zones in both the template and the image block, and performing cross-correlation for each corresponding sub-zone.

161. The system as in claim 158 further comprising the step of computing the overall cross-correlation score for the image block if the cross-correlation score for each sub-zone exceeds a threshold.

162. The system as in claim 146 further comprising means for identifying the locations that give locally maximum cross-correlation score.

163. A system for locating first and second objects in an image each object having substantially the same physical characteristics, and the ratio of the distance between the first and second objects and the size of each object is substantially invariant, the method comprising the steps of:

(a) means for displaying a digital image on a graphical user interface;

(b) means for estimating an approximate position of each object for forming initial input positions based on a format of the image;

(c) means for determining an estimate size of each object based on the distance obtained from the initial input positions of the first and second objects;

(d) means for forming a first searching window for the first object;

the center of the first window is determined by the input position of the first object and the size of the first window is determined by the estimate size of the first object;

(e) means for reiteratively positioning a template on the first search window for determining a location that gives a desired match of the first object relative to the template;

(f) means for forming a second searching window for the second object;

the center of the second window is determined by the input position of the second object and the size of the second window is determined by the estimate size of the second object; and

(g) means for reiteratively positioning a template on the second search window for determining a location that gives a desired match of the second object relative to the template.

164. The system as in claim 163, further comprising (h) means for determining a plurality of locations in each search window that give a desired matching response to the template.

165. The system as in claim 164 further comprising means for finding the best pair of locations from all possible combinations of identified locations of both the first and second objects.

166. The system as in claim 165, wherein the means for finding best pair of locations includes means for computing a plurality of figures of merit and means for determining a pair of locations that give a best combined figure of merit.

167. The system as in claim 166, wherein the means for computing a plurality of figures of merit includes means for computing individually or in combination an orientation, proximity, combined correlation score, proportion, profile, symmetry or evidence of mouth.

168. The system as in claim 167, wherein the first and second objects are first and second human eyes and wherein the means for computing an orientation includes means for measuring the difference between the orientation of the line connecting the first and second eyes, and the average orientation of the first and second eyes.

169. The system as in claim 167, wherein the means for computing a proximity includes measuring proximity of the first and second objects to the initial input positions, respectively.

170. The system as in claim 167, wherein computing a combined correlation score includes summing up the individual correlation scores obtained from the matching obtained in (e).

171. The system as in claim 167, wherein the first and second objects are first and second human eyes and wherein the means for computing a proportion includes means for computing a distance between the first and second eyes for satisfying an anthropological prior model.

172. The system as in claim 167, wherein the first and second objects are first and second human eyes and wherein the means for computing the profile includes means for pre-determining a model of the eye-to-eye profile and means for determining an actual profile from the image and means for computing a goodness-of-fit between the actual profile and the model profile.

173. The system as in claim 167, wherein the first and second objects are first and second human eyes and wherein the means for computing the symmetry includes means for computing a symmetry between first and second halves of a face window determined by the eye locations.

174. The system as in claim 167, wherein the first and second objects are first and second human eyes and wherein the means for computing the evidence of mouth includes means for computing strength and orientation of edges within a mouth window determined by the eye locations.

175. The system as in claim 164 further comprising means for normalizing the template for maximizing robustness of matching.

176. The system as in claim 164 further comprising means for creating a image block of the same size and orientation as the template.

177. The system as in claim 176 further comprising means for normalizing the image block for maximizing robustness of matching.

178. The system as in claim 176 further comprising means for extracting individual sub-zones in both the template and the image block, and performing cross-correlation for each corresponding sub-zone.

179. The system as in claim 176 further comprising means for computing the overall cross-correlation score for the image block if the cross-correlation score for each sub-zone exceeds a threshold.

180. The system as in claim 164 further comprising the step of identifying the locations that give locally maximum cross-correlation score.

181. A system for locating first and second objects each having substantially the same physical characteristics, and the ratio of the distance between the first and second objects and the size of each object is substantially invariant, the system comprising:

(a) means for displaying a digital image on a graphical user interface;

(b) means for estimating an approximate position of each object for forming initial input positions;

(c) means for determining an estimate orientation of each object based on an angle derived from the initial input positions of the first and second objects;

(d) means for forming a first searching window for the first object;

the center of the first window is determined by the initial input position of the first object and the size of the first window is determined by the estimate size of the first object;

(e) means for reiteratively positioning a template on the first search window for determining a location that gives a desired match of the first object relative to the template; and

(f) means for forming a second searching window for the first object;

the center of the second window is determined by the input position of the first object and the size of the second window is determined by the estimate size of the first object;

(g) means for reiteratively positioning a template on the second search window for determining a location that gives a desired match of the first object relative to the template.

182. The system as in claim 181 further comprising means for determining a plurality of locations that give a desired matching response to the template.

183. The system as in claim 181 further comprising the step of finding the best pair of locations from all possible combinations of identified locations of both the first and second objects.

184. The system as in claim 183, wherein the means for finding best pair of locations includes means for computing a plurality of figures of merit and determining a pair of locations that give a best combined figure of merit.

185. The system as in claim 184, wherein the means for computing a plurality of figures of merit includes means for computing individually or in combination an orientation, proximity, combined correlation score, proportion, profile, symmetry or evidence of mouth.

186. The method as in claim 185, wherein the first and second objects are first and second human eyes and wherein the means for computing an orientation includes means for measuring the difference between the orientation of the line connecting the first and second eyes, and the average orientation of the first and second eyes.

187. The system as in claim 185, wherein the means for computing a proximity includes means for measuring proximity of the first and second objects to the initial input positions, respectively.

188. The system as in claim 185, wherein the means for computing a combined correlation score includes means for summing up the individual correlation scores obtained from the matching obtained in (e).

189. The system as in claim 185, wherein the first and second objects are first and second human eyes and wherein the means for computing a proportion includes means for computing a distance between the first and second eyes for satisfying an anthropological prior model.

190. The system as in claim 185, wherein the first and second objects are first and second human eyes and wherein the means for computing the profile includes means for pre-determining a model of the eye-to-eye profile and means for determining an actual profile from the image and computing a goodness-of-fit between the actual profile and the model profile.

191. The system as in claim 185, wherein the first and second objects are first and second human eyes and wherein the means for computing the symmetry includes means for computing a symmetry between first and second halves of a face window determined by the eye locations.

192. The system as in claim 185, wherein the first and second objects are first and second human eyes and wherein the means for computing the evidence of mouth includes means for computing strength and orientation of edges within a mouth window determined by the eye locations.

193. The system as in claim 182 further comprising means for normalizing the template for maximizing robustness of matching.

194. The system as in claim 182 further comprising means for creating a image block of the same size and orientation as the template.

195. The system as in claim 194 further comprising means for normalizing the image block for maximizing robustness of matching.

196. The system as in claim 195 further comprising means for extracting individual sub-zones in both the template and the image block, and performing cross-correlation for each corresponding sub-zone.

197. The system as in claim 195 further comprising means for computing the overall cross-correlation score for the image block if the cross-correlation score for each sub-zone exceeds a threshold.

198. The system as in claim 183 further comprising means for identifying the locations that give locally maximum cross-correlation score.

199. A system for locating first and second objects in an image each object having substantially the same physical characteristics, and the ratio of the distance between the first and second objects and the size of each object is substantially invariant, the system comprising:

(a) means for displaying a digital image on a graphical user interface;

(b) means for estimating an approximate position of each object for forming initial input positions based on a format of the image;

(c) means for determining an estimate orientation of each object based on an angle derived from the initial input positions of the first and second objects;

(d) means for determining an estimate size of each object based on the distance obtained from the initial input positions of the first and second objects;

(e) means for forming a searching window for the first object;

the center of the window is determined by the input position of the first object and the size of the window is determined by the estimate size of the first object;

(f) means for reiteratively positioning a template on the search window for determining a location that gives a desired match of the first object relative to the template; and

(g) means for repeating steps (e) and (f) for the second object.

200. The system as in claim 199, further comprising (h) means for determining a plurality of locations in each search window that give a desired matching response to the template.

201. The system as in claim 200 further comprising means for finding the best pair of locations from all possible combinations of identified locations of both the first and second objects.

202. The system as in claim 201, wherein the means for finding best pair of locations includes means for computing a plurality of figures of merit and determining a pair of locations that give a best combined figure of merit.

203. The system as in claim 201, wherein the means for computing a plurality of figures of merit includes means for computing individually or in combination an orientation, proximity, combined correlation score, proportion, profile, symmetry or evidence of mouth.

204. The system as in claim 203, wherein the first and second objects are first and second human eyes and wherein the means for computing an orientation includes means for measuring the difference between the orientation of the line connecting the first and second eyes, and the average orientation of the first and second eyes.

205. The system as in claim 203, wherein the means for computing a proximity includes means for measuring proximity of the first and second objects to the initial input positions, respectively.

206. The system as in claim 203, wherein the means for computing a combined correlation score includes means for summing up the individual correlation scores obtained from the matching obtained in (e).

207. The system as in claim 203, wherein the first and second objects are first and second human eyes and wherein the means for computing a proportion includes means for computing a distance between the first and second eyes for satisfying an anthropological prior model.

208. The system as in claim 203, wherein the first and second objects are first and second human eyes and wherein the means for computing the profile includes means for pre-determining a model of the eye-to-eye profile and means for determining an actual profile from the image and computing a goodness-of-fit between the actual profile and the model profile.

209. The system as in claim 203, wherein the first and second objects are first and second human eyes and wherein the means for computing the symmetry includes means for computing a symmetry between first and second halves of a face window determined by the eye locations.

210. The system in claim 203, wherein the first and second objects are first and second human eyes and wherein the means for computing the evidence of mouth includes means for computing strength and orientation of edges within a mouth window determined by the eye locations.

211. The system as in claim 200 further comprising means for normalizing the template for maximizing robustness of matching.

212. The system as in claim 200 further comprising means for creating a image block of the same size and orientation as the template.

213. The system as in claim 212 further comprising means for normalizing the image block for maximizing robustness of matching.

214. The system as in claim 212 further comprising means for extracting individual sub-zones in both the template and the image block, and performing cross-correlation for each corresponding sub-zone.

215. The system as in claim 212 further comprising means for computing the overall cross-correlation score for the image block if the cross-correlation score for each sub-zone exceeds a threshold.

216. The system as in claim 200 further comprising means for identifying the locations that give locally maximum cross-correlation score.