Finding Color Regions

When searching the ball, obstacles or goals, we search for all pixel-connected regions whose pixels satisfy a predefined color check-mask. Once we have found an object, we only search within a small rectangular region centered at the object in consecutive camera images. To segement regions, we use a fast region growing algorithm. It starts with a seed pixel and consecutively adds pixels (in the directions up, left, down, right) to the region until the color specification is not satisfied any more. To avoid the algorithm to stop growing because of the presence of noise, that may separate regions which belong together, the algorithm only stops growing when more than $k=3$ consecutive pixels have been found that violate the color specification.