Segmentation for Image Semantic Analysis Based on Improved JSEG Algorithm
Zhou Yu · Video Engineering · 2014
JSEG algorithm is a classic image segmentation algorithm which is used in image processing widely,but the over-segmentation problem is so severely that this algorithm isn't suitable for Image Semantic Analysis.Two improvements to solve this issue are proposed.One is improved the initial seed point selection method,the other is consider the information of texture feature in the region merging process.The experiments on large amounts of images show that the proposed method reduces the over-segmentation phenomenon effective and conforms to the requirements of image semantic analysis.