Gradient based method for providing values for unknown pixels in a digital image
First Claim
1. A method for correcting one or more defect pixels in a defect region of a source image, which image has both the defect pixels and non-defect pixels each of which is represented by at least one defect pixel signal and non-defect pixel signal, respectively, the method comprising:
- a) searching for edges in a non-defect region of the image adjacent to the defect region by;
(i) comparing relative pixel parameters in the non-defect region with a predetermined tolerance to locate in the non-defect region a series of adjacent pixels representing an edge segment the members of which have relative parameters within the tolerance and which series is within a predetermined distance of the defect region; and
(ii) comparing the of the edge segment within a minimum length to identify as an edge a segment with a least the minimum length;
b) generating at least one pixel sign for the defect region based on the signal of at least one non-defect pixel lying in a direction from the defect pixel which is determined by the direction of an edge; and
c) replacing the defect pixel signal with the generated pixel signal.
4 Assignments
0 Petitions
Accused Products
Abstract
A method of correcting one or more defect pixels in a defect region of a source image, which image has both the defect pixels and non-defect pixels each of which is represented by at least one defect pixel signal and non-defect pixel signal, respectively. The method comprises:
a) searching for edges in a non-defect region of the image adjacent to the defect region by: (i) comparing relative pixel parameters in the non-defect region with a predetermined tolerance to locate a sequence of pixels representing an edge segment the members of which have relative parameters within the tolerance and which sequence is within a predetermined distance of the defect region; and (ii) comparing the length of an edge segment with a minimum length to identify as an edge a segment with at least the minimum length;
b) generating at least one pixel signal for the defect region based on the signal of at least one non-defect pixel lying in a direction from the defect pixel which is determined by the direction of an edge.
An apparatus for performing the method and a computer usable medium containing computer code for performing the method, are also provided.
140 Citations
41 Claims
-
1. A method for correcting one or more defect pixels in a defect region of a source image, which image has both the defect pixels and non-defect pixels each of which is represented by at least one defect pixel signal and non-defect pixel signal, respectively, the method comprising:
-
a) searching for edges in a non-defect region of the image adjacent to the defect region by;
(i) comparing relative pixel parameters in the non-defect region with a predetermined tolerance to locate in the non-defect region a series of adjacent pixels representing an edge segment the members of which have relative parameters within the tolerance and which series is within a predetermined distance of the defect region; and
(ii) comparing the of the edge segment within a minimum length to identify as an edge a segment with a least the minimum length;b) generating at least one pixel sign for the defect region based on the signal of at least one non-defect pixel lying in a direction from the defect pixel which is determined by the direction of an edge; and c) replacing the defect pixel signal with the generated pixel signal. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12)
-
-
13. A method of correcting one or more defect pixels in a defect region of a source image, which image has both the defect pixels and non-defect pixels each of which is represented by at least one defect pixel signal and non-defect pixel signal, respectively, the method comprising:
-
a) searching for edges in a non-defect region of the image adjacent to the defect region by;
(i) generating a gradient field for the non-defect region;
(ii) comparing gradients of the pixels to locate a start pixel in the non-defect region which has a gradient magnitude exceeding a preselected threshold;
(iii) comparing at least relative directional gradient attributes in the non-defect region with a predetermined tolerance to locate in the non-defect region series of adjacent pixels each series representing an edge segment, each series extending in at least one direction from the start pixel, the members of which have at least relative directional gradient attributes within the predetermined tolerance and which series is within a predetermined distance of the defect region;
(iv) comparing the length of the edge segments with a minimum length to identify as edges those edge segments with at least the minimum length;b) if at least two edges are found, comparing their relative directions and magnitudes with predetermined tolerances; c) generating pixel signals for the defect region between respective ends of two edges which have relative directions and magnitudes within the predetermined tolerances, which signals are based on at least one pixel signal of the two edges; and d) replacing defect pixel signals with the generated pixel signals. - View Dependent Claims (14)
-
-
15. A method of correcting one or more defect pixels in a defect region of a source image, which image has both the defect pixels and non-defect pixels each of which is represented by at least one defect pixel signal and non-defect pixel signal, respectively, the method comprising:
-
a) searching a non-defect region of the image adjacent to the defect region to locate in the non-defect region series of adjacent pixels each series representing an edge segment, and comparing the length of each segment with a predetermined minimum length to locate as edges those segments of at least the minimum length; b) searching for a geometric primitive by comparing at least one geometric parameter of a set of the edges with a geometric constraint corresponding to a geometric primitive, a geometric primitive being detected when the set at least meets the geometric constraint; c) generating pixel signals for defect region pixels each based on the signal of at least one non-defect pixel lying on a predictor line having a direction which has been-determined by the direction of at least one part of a detected geometric primitive; and d) replacing defect pixel signals with the generated pixel signals. - View Dependent Claims (16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32)
-
-
33. A method of correcting one or more defect pixels in a defect region of a source image, which image has both the defect pixels and non-defect pixels each of which is represented by at least one defect pixel signal and non-defect pixel signal, respectively, the method comprising:
-
a) searching a non-defect region of the image adjacent to the defect region for sequences of pixels representing edge segments, and comparing the length of each segment with a predetermined minimum length to locate as edges those segments of at least the minimum length the locations of the edges together defining an edge map; b) searching for a geometric primitive by comparing at least one geometric parameter of a set of edges with a geometric constraint corresponding to a geometric primitive, a geometric primitive being detected when the set at least meets the geometric constraint; c) generating pixel signals for defect region pixels each based on the signal of at least one non-defect pixel lying on a predictor line having a direction determined by the direction of at least one part of a detected geometric primitive; and d) replacing defect pixel signals with the generated pixel signals; wherein, prior to searching for a geometric primitive; (i) for at least one edge intersecting the defect region, the direction of a defect part of that edge is determined based on the direction of the edge; (ii) searching a non-defect region of the image adjacent to the defect region for a sequence of adjacent pixels in the non-defect region representing an edge segment extending from a first non-defect pixel intersected by the defect part of the edge, and comparing the length of such a sequence with a predetermined minimum length to locate as a defect exit part of the edge a sequence of at least the minimum length; and (iii) when a defect exit part is located, the at least one edge, its corresponding defect part and defect exit part are added to the edge map as a connected edge, or when a defect exit part is not located the edge and its defect part are added to the edge map as a not connected edge; and wherein, prior to searching for a geometric primitive, an edge thinning is performed on the edges in the edge map by; (i) locating a set of matched edges, where edges match when they are within a predetermined angle tolerance and distance of a selected one of the edges of the set; (ii) selecting the best edge of the set which is the strongest connected edge, if any, or if none, is the strongest not connected edge of the set, where the strength of the edge is defined by the average gradient magnitude of the pixels of the edge, and deleting the remainder of the edges of the set from the edge map. - View Dependent Claims (34, 35)
-
-
36. A method of correcting one or more defect pixels in a defect region of a source image, which image has both the defect pixels and non-defect pixels each of which is represented by at least one defect pixel signal and non-defect pixel signal, respectively, the method comprising:
-
a) searching a non-defect region of the image adjacent to the defect region for sequences of pixels representing edge segments, and comparing the length of each segment with a predetermined minimum length to locate as edges those segments of at least the minimum length; b) searching for a geometric primitive by comparing at least one geometric parameter of a set of the edges with a geometric constraint corresponding to a geometric primitive, a geometric primitive being detected when the set at least meets the geometric constraint; c) generating pixel signals for defect region pixels each based on the signal of at least one non-defect pixel lying on a predictor line having a direction determined by the direction of at least one part of a detected geometric primitive; and d) replacing defect pixel signals with the generated pixel signals; wherein the locations of the edges together define an edge map and wherein, prior to generating pixel signals, a detected geometric primitive is extracted by extending or reducing, in the map, one or more edges of the detected primitive so that they intersect in the same pattern as the primitive, or by replacing in the map the edges detected as a primitive with one or more new edges representing the primitive; and wherein the signal generated for a defect pixel within a predetermined distance of a plurality of neighboring parts of a detected geometric primitive in the edge map is determined by; first determining estimated pixel signals for the defect pixel based on at least one non-defect pixel lying on each of the predictor lines passing through the defect pixel, and then determining a final pixel signal which is a weighted sum of estimated pixel signals, each estimated pixel signal being weighted in an inverse relationship to the distance between the predictor line and the corresponding geometric primitive part.
-
-
37. An apparatus for correcting one or more defect pixels in a defect region of a source image, which image has both the defect pixels and non-defect pixels each of which is represented by at least one defect pixel signal and non-defect pixel signal, respectively, the apparatus comprising a programmed computer with a memory for storing the source image, the programmed computer having:
-
a) means for searching for edges in a non-defect region of the image adjacent to the defect region by;
(i) comparing relative pixel parameters in the non-defect region with a predetermined tolerance to locate in the non-defect region a series of adjacent pixels representing an edge segment the members of which have relative parameters within the tolerance and which series is within a predetermined distance of the defect region; and
(ii) comparing the length of an edge segment with a minimum length to identify as an edge a segment with at least the minimum length;b) means for generating at least one pixel signal for the defect region based on the signal of at least one non-defect pixel lying in a direction from the defect pixel which is determined by the direction of the edge; and c) means for replacing a defect pixel signal with the generated pixel signal.
-
-
38. An apparatus for correcting one or more defect pixels in a defect region of a source image, which image has both the defect pixels and non-defect pixels each of which is represented by at least one defect pixel signal and non-defect pixel signal, respectively, comprising a programmed computer with a memory for storing the source image, the programmed computer having:
-
a) means for searching a non-defect region of the image adjacent to the defect region to locate in the non-defect region series of adjacent pixels each series representing an edge segment, and comparing the length of each segment with a predetermined minimum length to locate as edges those segments of at least the minimum length; b) means for searching for a geometric primitive by comparing at least one geometric parameter of a set of edges with a geometric constraint corresponding to a geometric primitive, a geometric primitive being detected when the set at least meets the geometric constraint; c) means for generating pixel signals for defect region pixels each based on the signal of at least one non-defect pixel lying on a predictor line having the direction determined by the direction of at least one edge of a detected geometric primitive; and d) means for replacing defect pixel signals with the generated pixel signals.
-
-
39. A computer program product, comprising a computer readable storage medium including program means for correcting one or more defect pixels in a defect region of a source image, which image has both the defect pixels and non-defect pixels each of which is represented by at least one defect pixel signal and non-defect pixel signal, respectively, the program means comprising:
-
a) means for searching for edges in a non-defect region of the image adjacent to the defect region by;
(i) comparing relative pixel parameters in the non-defect region with a predetermined tolerance to locate in the non-defect region a series of adjacent pixels representing an edge segment the members of which have relative parameters within the tolerance and which series is within a predetermined distance of the defect region; and
(ii) comparing the length of an edge segment with a minimum length to identify as an edge a segment with at least the minimum length;b) means for generating at least one pixel signal for the defect region based on the signal of at least one non-defect pixel lying in a direction from the defect pixel which is determined by the direction of the edge; and c) means for replacing the defect pixel signal with the generated pixel signal.
-
-
40. A computer program product, comprising a computer readable storage medium with program means for correcting one or more defect pixels in a defect region of a source image, which image has both the defect pixels and non-defect pixels each of which is represented by at least one defect pixel signal and non-defect pixel signal, respectively, the program means comprising:
-
a) means for searching a non-defect region of the image adjacent to the defect region to locate in the non-defect region series of adjacent pixels each representing an edge segment, and comparing the length of each segment with a predetermined minimum length to locate as edges those segments of at least the minimum length; b) means for searching for a geometric primitive by comparing at least one geometric parameter of a set of edges with a geometric constraint corresponding to a geometric primitive, a geometric primitive being detected when the set at least meets the geometric constraint; c) means for generating pixel signals for defect region pixels each based on the signal of at least one non-defect pixel lying on a predictor line having a direction determined by the direction of at least one edge of a detected geometric primitive; and d) means for replacing defect pixels signal with the generated pixel signals.
-
-
41. A method of correcting at least one defect pixel in a defect region of a source image, which image has both the defect pixel and non-defect pixels each of which is represented by at least one defect pixel signal and non-defect pixel signal, respectively, the method comprising:
-
allocating a plurality of line segments through a selected defect pixel, wherein each said line segment is composed of the defect pixel and non-defect pixels about said selected defect pixel; determining at least one representative model of the non-defect pixel signals along each said line segment; determining the deviation of the non-defect pixel signals along each said line segment from the model corresponding to the line segment; determining at least one dominant direction of the local gradient field; selecting at least one line segment based on a figure of merit; calculating an estimated pixel signal for the defect pixel based on at least one non-defect pixel for each selected line segment; generating a final pixel signal which is a weighted sum of estimated pixel signals with weights proportional to figures of merit of the selected line segments, wherein the figure of merit is a value which increases with decreased deviation of the non-defect pixels along the selected line segment, and increases with the alignment of the line segment with the dominant directions of the local gradient field and replacing the defect pixel signal with the generated pixel signal.
-
Specification