Color image segmentation by neutrosophy approach

In this thesis, neutrosophic grey level image segmentation has been examined. Although this method is pretty good at grey level image segmentation, it has some deficiencies. This new approach obtained by eliminating these deficiencies has been used for image segmentation. First, by turning color images to be segmentated into L, a, b color space, each color channel is turned to neutrosophic domain as independent. To calculate indefinite membership of neutrosophic domain, local entropy filter has been used different from previous methods. ? median and ß enhancement are used for modifying segmentation algorithm. For each image, a new entropy-based method has been used to determine ? and ß. Consequently, each color channel in NS domain is segmentated by using ?-median clustering method.This approach developed has been applied to real images, unreal images and noised real images. Segmentation results have been compared with the results which are generated by mean shift and Waveseg algorithm to evaluate the performance of this newly used algorithm. Obtained results showed that success of proposed method is good.