Method and a device for filling occluded areas of a depth or disparity map estimated from at least two images
A method of filling occluded areas of a depth or disparity map estimated from at least two images and consisting of a matrix of pixels forming a set of lines and columns, each pixel of the map being associated with a depth or disparity value (pixel value) and any pixel of an occluded area (invalid pixel) being associated with a pixel value identifiable as being invalid. This method comprises a step of processing each line of the map pixel by pixel in a predefined direction and assigning each invalid pixel Pi encountered on a current line a value determined as a function of the values associated with the pixels in a predefined vicinity around the first valid pixel P 1 , if any, that follows the invalid pixel Pi on the current line and the value V(P 1 ) of which corresponds to a greater depth or a lesser disparity relative to the value V(P 0 ) of the last valid pixel P 0 , if any, that precedes the invalid pixel Pi on the current line.
1. A method of filling occluded areas of a depth or disparity map estimated from at least two images and consisting of a matrix of pixels forming a set of lines and columns, each pixel of the map being associated with a depth or disparity value, called pixel value, and any pixel of an occluded area, called invalid pixel, being associated with a pixel value identifiable as being invalid, said method comprising:
a step of processing each line of the map, pixel by pixel, in a predefined direction; and
a step of assigning a pixel value to each invalid pixel Pi encountered on a current line, said pixel value being determined as a function of the values associated with the pixels in a predefined vicinity around a chosen pixel or being a predefined default value, comprising
if there is a last valid pixel P 0 that precedes the invalid pixel Pi on the current line, and there are one or more valid pixels that follow the invalid pixel Pi on the current line, searching among said valid pixels that follow the invalid pixel Pi for a first valid pixel P 1 the value V(P 1 ) of which corresponds to a greater depth or a lesser disparity relative to the value V(P 0 ) of the last valid pixel P 0 , and if said pixel P 1 is found choosing said pixel P 1 as said chosen pixel;
if there is a valid pixel P 0 and there are one or more said valid pixels that follow the invalid pixel Pi but there is no valid pixel P 1 , choosing the first valid pixel P 2 that follows the invalid pixel Pi as said chosen pixel,
if there is no valid pixel P 0 but there are one or more said valid pixels that follow the invalid pixel Pi, choosing the first valid pixel P 2 that follows the invalid pixel Pi as said chosen pixel,
if there is a valid pixel P 0 but there is no valid pixel P 2 , choosing said last valid pixel P 0 as said chosen pixel, and
said assigned pixel value being a predefined default value Vd if there is no valid pixel P 0 and no valid pixel P 2 .
2. The method according to claim 1 , wherein the pixel value assigned to the invalid pixel Pi is determined as a function of the values associated with the pixels in the vicinity of a pixel concerned by application of the median or mean mathematical function to said values of pixels in the vicinity concerned.
3. The method according to claim 1 , wherein a vicinity around a pixel concerned consists of a block of valid pixels centered on the pixel concerned.
4. The method according to claim 1 , wherein said depth or disparity map is obtained from a pair of right-hand and left-hand images and wherein the predefined direction of processing a row of pixels of the map corresponds to the direction from right to left if the map is estimated from the left-hand image and to the direction from left to right if the map is estimated from the right-hand image.
5. A method of digitally processing a depth or disparity map, comprising an operation of filling occluded areas using the method according to claim 1 .
6. The processing method according to claim 5 , wherein the operation of filling the occluded areas is preceded by an operation of closure morphological filtering eliminating estimation errors in occluded areas of the map.
7. The processing method according to claim 5 , wherein the operation of filling the occluded areas is preceded by an operation of local uniformization of the map using over the whole of the map median filtering of the values of the pixels analyzed per block of predefined size.
8. A device for digitally processing a depth or disparity map, comprising a module for filling occluded areas of the map adapted to use the filling method according to claim 1 .
9. The device according to claim 8 , further comprising a closure morphological filter module for eliminating estimation errors in occluded areas of the map before the filling of the occluded areas.
10. The device according to claim 8 , further comprising a module for local uniformization of the map using over the whole of the map median filtering of the pixel values analyzed per block of predefined size before the filling of the occluded areas.
11. A non-transitory computer readable medium having instructions recorded thereon, the instructions being adapted to execute the method according to claim 1 when executed in a computer system.