Effect of color space on color image segmentation

![CDATA[A study of color image segmentation with its dependence on color space representation is presented. Segmentation has been one of the basic procedures in image processing. Because of the three-fold increase in color signal dimension as compared to black-and-white images, an advantage resulting from the choice of color space representation could be taken to enhance the performances of processes such as segmentation and feature matching. However, the choice of a particular color space is still largely application dependent. This work attempts to study a number of popular color space schemes on the basis of the maximum information that the space is able to convey to the segmentation process. Thus, a reduction in the complexity of the segmentation procedure is achievable when it is operating on a single color space domain. The amount of information contained in the segmented objects is adopted as a measure to determine the segmentation rule. Several aerial images over planted fields are employed in experiments and their satisfactory segmentation results are used to conclude the study.]]