Image processing device for processing grey level images

• US 5,625,717 A
• Filed: 06/22/1993
• Issued: 04/29/1997
• Est. Priority Date: 06/24/1992
• Status: Expired due to Term

• Current Assignee: Mitsubishi Denki Kabushiki Kaisha (Mitsubishi Electric Corporation)
• Original Assignee: Mitsubishi Denki Kabushiki Kaisha (Mitsubishi Electric Corporation)
• Inventors: Hashimoto, Manabu, Sumi, Kazuhiko
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): KELLEY, CHRISTOPHER S
• Application Number:
  US08/081,463
• Time in Patent Office: 1,407 Days
• Field of Search: 382/50, 382/22, 382/31, 382/41, 382/48, 382/54, 382/254, 382/256, 382/258, 382/260, 382/266, 382/263, 382/270, 382/274
• US Class Current: 382/260
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/136 involving thresholding
• CPC Class Codes: G06V 10/26 Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
• CPC Class Codes: G06V 10/28 Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns

Abstract

An image processing device includes: a DOG filtering means for applying a DOG filter upon the grey level image of an object; and a tri-level thresholding means for thresholding the output of the DOG filtering means at two distinct threshold levels, thereby obtaining a tri-level image of the object. The 0-level regions of the tri-level image correspond to regions of the original image, such as the surface areas of the object, within which the variation of the grey level is small. The boundary between two 0-regions is marked by adjacent parallel strips of a (+)-region and a (-)-region, wherein the zero-crossing line between the (+)-region and the (-)-region clearly defines the contour line of the 0-regions. This method of tri-level quantization provides a basis for efficient template matching of the images of objects, recognition or identification of objects, the detection of movement vectors of objects, and the detection of the parallax of a continuously moving objects.

Claims

1. An image processing device comprising:

original image input means for inputting a grey level signal representing a grey level image of an object;

spatial band-pass filter means, coupled to said original image input means for producing a filtered signal in which substantially uniform regions of the grey level image are made substantially zero in first corresponding portions of the filtered signal, and transitions between substantially uniform regions are preserved second corresponding portions of the filtered signal; and

tri-level thresholding means, coupled to said spatial band-pass filter means, for thresholding said filtered signal at two distinct threshold levels to generate a tri-level image signal identifying substantially uniform regions of the grey level signal as those regions of the grey level signal corresponding to regions of the filtered signal lying between the threshold levels and transitions as regions of the grey level signal corresponding to a pair of adjacent excursions of the filtered signal away from substantially zero, a first excursion passing beyond one of the two distinct thresholds, and a second excursion passing beyond another of the two distinct thresholds.

13. The device of claim 1, wherein the original input means includes a camera.

21. The device of claim 1, wherein one of the two distinct threshold values has a positive threshold value, and the other of the two distinct threshold levels has a negative threshold value.

23. The device of claim 1, wherein the input signals are received as image frames, and wherein the filter means and tri-level thresholding means filter and threshold each image frame.

2. An image processing device comprising:

original image input means for inputting an image signal representing a grey level image of an object;

reference pattern supply means for supplying a reference signal representing a grey level two-dimensional reference pattern;

spatial band-pass filter means, coupled to said original image input means and said reference pattern supply means, for applying a spatial band-pass filter upon said reference signal and said image signal to generate a filtered reference signal in which substantially uniform regions of the grey level reference pattern are made substantially a reference value in first corresponding portions of the filtered reference signal, and transitions between substantially uniform regions are preserved in second corresponding portions of the filtered reference signal and a filtered image signal in which substantially uniform regions of the grey level image are made substantially a reference value in first corresponding portions of the filtered image signal, and transitions between substantially uniform regions are preserved in second corresponding portions of the filtered image signal;

tri-level thresholding means, coupled to said spatial band-pass filter means, for thresholding said filtered reference signal and said filtered image signal at two distinct threshold levels to generate a tri-level reference signal and a tri-level image signal, respectively, each tri-level signal identifying substantially uniform regions of a corresponding grey level signal as those regions of the corresponding grey level signal corresponding to regions of the filtered signal lying between the threshold levels and transitions as regions of the corresponding grey level signal corresponding to a pair of adjacent excursions of the filtered signal away from substantially zero, a first excursion passing beyond one of the two distinct thresholds and a second excursion passing beyond another of the two distinct thresholds;

similarity calculator means, coupled to said tri-level thresholding means, for determining a portion of said tri-level image of said object which is in best agreement with said tri-level reference pattern, wherein said similarity calculator means includes;

means for setting a relative displacement of said tri-level image of said object with respect to said tri-level reference pattern;

means for calculating a similarity of said tri-level image of said object and said tri-level reference pattern at said relative displacement;

means for varying said relative displacement successively within a predetermined range; and

means for determining a relative displacement within said predetermined range at which said similarity reaches a maximum.

3. An image processing device as claimed in claim 2, wherein said similarity is calculated by said calculating means in accordance with an equation:

space="preserve" listing-type="equation">M(a,b)=Σ

Σ

f(R(i,j),I(i+a,j+b))

wherein M(a,b) is said similarity at relative displacement (a,b);

R(i,j) is said tri-level reference pattern;

I(i,j) is said tri-level object image; and

summation is taken over 0≧

i≧

n and 0≧

j≧

m, m being a size of said reference pattern in a direction of index j, and n being a size thereof in a direction of index i, f(α

,β

) being an evaluation function.

14. The device of claim 2, wherein the original Input means includes a camera.

22. The device of claim 2, wherein one of the two distinct threshold values has a positive threshold value, and the other of the two distinct threshold levels has a negative threshold value.

24. The device of claim 2, wherein the input signals are received as image frames, and wherein the filter means and tri-level thresholding means filter and threshold each image frame.

4. An image processing device comprising:

object image input means for inputting a signal representing a grey level image of an object;

spatial band-pass filter means, coupled to said object image input means, for applying a spatial band-pass filter upon said signal to generate a filtered image signal representing a filtered image of said object in which substantially uniform regions of the grey level image are made substantially a reference value in first corresponding portions of the filtered image signal, and transitions between substantially uniform regions are preserved in second corresponding portions of the filtered image signal;

tri-level thresholding means, coupled to said spatial band-pass filter means, for thresholding said filtered image signal at a first and a second threshold level, said first threshold level being greater than said second threshold level, said tri-level thresholding means generating a tri-level image signal representing a tri-level image of said object which includes;

first regions of the tri-level image corresponding to a pair of adjacent excursions of the filtered image signal, a first excursion passing above said first threshold level and a second excursion passing below said second threshold level, and

second regions of the tri-level image corresponding to regions of said filtered image of said object having values between said first and second threshold levels; and

regional segmentation means, coupled to said tri-level thresholding means, for segmenting said tri-level image of said object, wherein said regional segmentation means segments only said second regions of said tri-level image of said object.

15. The device of claim 4, wherein the original input means includes a camera.

5. An image processing device comprising:

object image input means for inputting a signal representing a grey level image of an object;

spatial band-pass filter means, coupled to said object image input means, for applying a spatial band-pass filter upon said signal to generate a filtered image signal representing a filtered image of said object in which substantially uniform regions of the grey level image are made substantially a reference value in first corresponding portions of the filtered image signal, and transitions between substantially uniform regions are preserved in second corresponding portions of the filtered image signal;

tri-level thresholding means, coupled to said spatial band-pass filter means, for thresholding said filtered image signal at a first and a second threshold level, said first threshold level being greater than said second threshold level, said tri-level thresholding means generating a tri-level image signal representing a tri-level image of said object which includes;

first regions of the tri-level image corresponding to a pair of adjacent excursions of the filtered image signal, a first excursion passing above said first threshold level and a second excursion passing below said second threshold level, and

second regions of the tri-level image corresponding to regions of said filtered image of said object having values between said first and second threshold levels;

segmentation means coupled to said tri-level thresholding means, including;

first segmentation means for segmenting said first regions corresponding to regions of said filtered image of said object below said second threshold level, said first segmentation means generating first segmented regions,

second segmentation means for segmenting said second regions corresponding to regions of said filtered image between said first and second threshold levels, said second segmentation means generating second segmented regions, and

third segmentation means for segmenting said third regions corresponding to regions of said filtered image of said object below said second threshold level, said third segmentation means generating third segmented regions; and

segment synthesizer means, coupled to said segmentation means, for synthesizing predetermined segmented regions among said first, second, and third segmented regions output from said segmentation means.

6. An image processing device as claimed in claim 5, wherein said first threshold level is negative and said second threshold level is positive, and said segment synthesizer means synthesizes each one of said second segmented regions with an adjacent first or third segmented region.

16. The device of claim 5, wherein the original input means includes a camera.

7. An image processing device comprising:

object image input means for inputting a signal representing a grey level digital image of an object;

spatial band-pass filter means, coupled to said object image input means, for applying a spatial band-pass filter upon said signal to generate a filtered image signal representing a filtered image of said object in which substantially uniform regions of the grey level image are made substantially a reference value in first corresponding portions of the filtered image signal, and transitions between substantially uniform regions are preserved in second corresponding portions of the filtered image signal;

tri-level thresholding means, coupled to said spatial band-pass filter means, for thresholding said filtered image signal at a first and a second threshold level, said first threshold level being greater than said second threshold level, said tri-level thresholding means generating a tri-level image signal representing a tri-level image of said object which includes;

first regions of the tri-level image corresponding to a pair of adjacent excursions of the filtered image signal, a first excursion passing above said first threshold level and a second excursion passing below said second threshold level, and

second regions of the tri-level image corresponding to regions of said filtered image of said object having values between said first and second threshold levels;

regional segmentation means, coupled to said tri-level thresholding means, for segmenting said tri-level image of said object, wherein said regional segmentation means segments only said second regions of said tri-level image of said object, said regional segmentation means generating segmented regions;

contour detector means, coupled to said spatial band-pass filter means, for extracting contour pixels from said filtered image of said object; and

contour segmentation means, coupled to said regional segmentation means and said contour detector means, for segmenting said contour pixels by determining a segmented region output from said regional segmentation means which is closest to each contour pixel, said contour segmentation means generating contour segments each of which includes contour pixels which are determined to be closest to a segmented region output relative to said regional segmentation means.

17. The device of claim 7, wherein the original Input means includes a camera.

8. An image processing device for detecting movement of an object from at least two successive images thereof, said image processing device comprising:

object image input means for inputting first and second signals representing successive grey level images of the object;

spatial band-pass filter means, coupled to said object image input means, for applying a spatial band-pass filter upon said first and said second signals to generate first and second filtered signals which represent respective first and second filtered images of the object in which substantially uniform regions of the successive grey level images are made substantially a reference value in first corresponding portions of the first and second filtered image signals, and transitions between substantially uniform regions are preserved in second corresponding portions of the first and second filtered image signals;

tri-level thresholding means, coupled to said spatial band-pass filter means, for thresholding each of said first and second filtered signals at two distinct threshold levels to generate first and second tri-level image signals, respectively, the first and second tri-level image signals identifying substantially uniform regions of the successive grey level signals as those regions of the successive grey level signal corresponding to regions of the first and second filtered signals, respectively, lying between the threshold levels and transitions as regions of the successive grey level signals corresponding to a pair of adjacent excursions of the first and second filtered signals, respectively, from substantially the reference value, a first excursion passing beyond one of the two distinct thresholds and a second excursion passing beyond another of the two distinct thresholds;

labeling means, coupled to said tri-level thresholding means, for labeling said first and said second tri-level images of the object, said labeling means generating labeled regions of said first and said second tri-level images;

label position detector means, coupled to said labeling means, for detecting a position of each labeled region of said first and said second tri-level images output from said labeling means; and

movement vector extractor means, coupled to said label position detector means, for extracting a movement vector of said object between said first and said second images, wherein said movement vector extractor means includes;

means for determining a position of a labeled region of said second tri-level image which is closest to a position of a labeled region of said first tri-level image; and

means for determining a vector connecting said position of said labeled region of said first tri-level image to said position of said labeled region of said second tri-level image.

9. An image processing device as claimed in claim 8, wherein said position of said labeled regions is a center of gravity of respective labeled regions.

18. The device of claim 8, wherein the original input means includes a camera.

10. An image processing device for detecting a parallax of an object from successive image frames thereof, said image processing device comprising:

object image input means for inputting signals representing successive grey level image frames of an object;

spatial band-pass filter means, coupled to said object image input means, for applying a spatial band-pass filter upon said signals representing said successive image frames of said object to generate filtered image signals representing filtered image frames of said object in which substantially uniform regions of the grey level image are made substantially a reference value in first corresponding portions of the filtered image signals, and transitions between substantially uniform regions are preserved in second corresponding portions of the filtered signals;

tri-level thresholding means, coupled to said spatial band-pass filter means, for thresholding said filtered signals at two distinct threshold levels to generate tri-level image signals, representing successive tri-level image frames, respectively, each tri-level image signal identifying substantially uniform regions of a corresponding grey level signal as those regions of the corresponding grey level signal corresponding to regions of the filtered signal lying between the threshold levels and transitions as regions of the corresponding grey level signal corresponding to a pair of adjacent excursions of the filtered signal array from substantially a reference value, a first excursion passing beyond one of the two distinct thresholds and a second excursion passing beyond another of the two distinct thresholds;

labeling means, coupled to said tri-level thresholding means, for labeling said tri-level image frames of said object, said labeling means generating labeled regions of said tri-level image frames;

label position detector means, coupled to said labeling means, for detecting a position of each labeled region of said tri-level image frames output from said labeling means;

label correspondence calculator means, coupled to said label position detector means, for determining correspondence among successive positions of said labeled regions of said tri-level image frames, wherein said label correspondence calculator means determines said correspondence by establishing a correspondence between a first position of a first labeled region of a first tri-level image frame and a second position of a second labeled region of a second tri-level image frame, said first and second tri-level image frame being two arbitrary successive tri-level image frames among said tri-level image frames output from said tri-level thresholding means, where said second position is closest to said first position among said positions of said labeled regions of said second tri-level image frame;

contour extractor means, coupled to said spatial band-pass filter means, for extracting contours from a first and a last filtered image frames output from said spatial band-pass filter means;

parallax calculator means, coupled to said label correspondence calculator means and said contour extractor means, for calculating a parallax of said object between said first and last filtered image frame, wherein said parallax calculator means includes;

means for determining a correspondence between said contours of said first and last filtered image frames; and

means for determining said parallax by determining a movement of said contours of said first and last filtered image frames on the basis of said correspondence established between said contours of said first and last filtered image frames.

11. An image processing device as claimed in claim 10, wherein said position of said labeled regions is a center of gravity of respective labeled regions.

19. The device of claim 10, wherein the original input means includes a camera.

12. An image processing device comprising:

original image input means for inputting a signal representing a grey level image of an object;

means for supplying model data;

spatial band-pass filter means, coupled to said original image input means, for applying a spatial band-pass filter upon said signal to generate a filtered image signal representing a filtered image of said object in which substantially uniform regions of the grey level image are made substantially a reference value in first corresponding portions of the filtered image signal, and transitions between substantially uniform regions are preserved in second corresponding portions of the filtered image signal;

tri-level thresholding means, coupled to said spatial band-pass filter means, for thresholding said filtered image signal at two distinct threshold levels to generate a tri-level image signal, identifying substantially uniform regions of the grey level image as those regions of the grey level image corresponding to regions of the filtered image signal lying between the threshold levels and transitions as regions of the grey level image corresponding to a pair of adjacent excursions of the filtered image signal away from substantially the reference value, a first excursion passing beyond one of the two distinct thresholds and a second excursion passing beyond another of the two distinct thresholds;

labeling means, coupled to said thresholding means, for labeling said tri-level image of said object, said labeling means generating labeled regions of said tri-level image;

grey level information extractor means, coupled to said original image input means and said labeling means, for extracting grey level information of said grey level image of said object with respect to each labeled region output from said labeling means;

grey leveled label generator means, coupled to said labeling means and said grey level information extractor means, for generating information upon grey-leveled labeled regions by which said grey level information with respect to each labeled region is added to information upon each labeled region; and

matching means, coupled to said model data and said grey leveled label generator means, for matching said information upon grey-leveled labeled regions with said model data, said matching means thereby generating a recognition information upon said object.

20. The device of claim 12, wherein the original input means includes a camera.