Method for controlling tracking using a color model, corresponding apparatus and non-transitory program storage device

1. A video tracker device comprising:

a memory for storing a video image comprising at least first and second frames, each pixel in the video image having at least one color component; and

a processor configured to:

obtain a window in the second frame of a video image corresponding to a window in the first frame of the video image using a tracking algorithm in a tracking mode;

compute a foreground color model for each one of at least two groups of pixels in the window in the first frame, each foreground color model comprising at least one color component computed from values of the color components of pixels in the corresponding group;

compute a background color model for each one of at least two groups of pixels in a background area around the window in the first frame, each background color model comprising at least one color component computed from values of the color components of pixels in the corresponding group;

determine a foreground color distance between each pixel in the window in the second frame and each of the foreground color models according to color component values and determining a foreground minimum color distance for the each pixel as a minimum of the determined foreground color distances;

determine a background color distance between each pixel in the window in the second frame and each of the background color models according to color component values and determining a background minimum color distance for the each pixel as a minimum of the determined background color distances;

assign a pixel confidence value for each pixel in the second frame according to the corresponding foreground and background minimum color distances for the each pixel in the window in the second frame;

assign a window confidence value for the window in the second frame according to the pixel confidence values of pixels in the window in the second frame;

if the window confidence value is greater than a first confidence threshold, selecting the tracking mode; and

if the window confidence value is not greater than the first confidence threshold, selecting a mode different from the tracking mode.

2. The video tracker according to claim 1 , wherein each color component of a color model of a group of pixels is computed by taking an average of corresponding components of pixels in that group.

3. The video tracker according to claim 1 , wherein the processor is further configured to cluster pixels in the window in the first frame into a first predefined number of clusters based on color component values, and select the at least two groups of pixels in the window in the first frame by selecting a subset of the first predefined number of clusters, each group being one of the clusters in the selected set.

4. The video tracker according to claim 1 , wherein the processor is further configured to assign a weight to each pixel in the window in the first frame, obtain a sum of weights of pixels in each of the first predefined number of clusters, and select the subset by selecting clusters in the first predefined number of clusters having corresponding sums over a first sum threshold.

5. The video tracker according to claim 1 , wherein the processor is further configured to assign a weight of a pixel in the window in the first frame by assigning the weight according to a spatial distance between the pixel and a center of the window in the first frame.

6. The video tracker according to claim 1 , wherein the processor is further configured to cluster pixels in the background area in the first frame into a third predefined number of clusters based on pixel values, and select the at least two groups of pixels in the background area in the first frame by selecting a subset of the third predefined number of clusters, each group being one of the clusters in the selected subset.

7. The video tracker according to claim 6 , wherein the processor is further configured to assign a weight to each pixel in the background area in the first frame, obtain a sum of weights of pixels in each of the third predefined number of clusters, and select the subset by selecting clusters in the third predefined number of clusters having corresponding sums over a second sum threshold.

8. A video tracker device comprising:

a memory for storing a video image comprising at least first and second frames, each pixel in the video image having at least one color component; and

a processor configured to:

obtain a window in a second frame of a video image corresponding to a window in a first frame of the video image using a tracking algorithm in a tracking mode, wherein each pixel in the video image has at least one color component;

compute a foreground color model for each one of at least two groups of pixels in the window in the first frame, each foreground color model comprising at least one color component computed from values of the color components of pixels in the corresponding group;

compute a background color model for each one of at least two groups of pixels in a background area around the window in the first frame, each background color model comprising at least one color component computed from values of the color components of pixels in the corresponding group;

determine a foreground color distance between a plurality of pixels in the window in the second frame and each of the foreground color models according to color component values and determining a foreground minimum color distance for each pixel of said plurality of pixels as a minimum of the determined foreground color distances;

determine a background color distance between a plurality of pixels in the window in the second frame and each of the background color models according to color component values and determining a background minimum color distance for each pixel of said plurality of pixels as a minimum of the determined background color distances;

assign a pixel confidence value for each pixel of said plurality of pixels in the second frame according to the corresponding foreground and background minimum color distances for the each pixel of said plurality of pixels in the window in the second frame;

select among the tracking mode or a mode different from the tracking mode, based on a window confidence value for the window in the second frame, wherein the window confidence value is obtained according to the pixel confidence values of said plurality of pixels in the window in the second frame.