Improved Coral Reef Images Segmentation using Modified JSEG Algorithm

Abstract

Underwater coral reef image segmentation suffers from various challenges due to various factors especially variation in illumination, different water turbidity, different water depth, variation in color, texture and shape of the coral reef species. In this paper, we modified an original automatic color image segmentation called JSEG to enable better coral reef segmentation process. The modification involves the substitution of General Lloyd Algorithm and agglomerative algorithm in the original JSEG version with the k-means algorithm. In addition, the newly modified JSEG algorithm process image in L*a*b color space to provide better processing of underwater image color property while k-means algorithm is used to segment the color within the specified cluster number. The experimental results showed that the modified JSEG algorithm could segment the coral reefs better than the original JSEG algorithm.