Abstract: JSEG algorithm is a method of segmentation of color images which can make the image filtering, color space quantization and spatial segmentation in due order. However, using the method directly in segmentation could result in inaccurate of edge due to fuzzy edge in the remote sensing image, or lead to excessive segmentation because of different shades in the region. To obtain the better effect of segmentation,the local homogeneous matrix which gives reasonably description for color homogeneity in region and the boundary of different region was used to correct the local J value of traditional JSEG algorithm in this paper. In addition, to weaken or to eliminate the over segmentation, LBP/C operator which gives stable description for the texture information of image was used to combine class maps with similar texture information. Experimental results show that the above methods work well.

Key words: High resolution color remote sensing image, JSEG algorithm, Local homogeneity, LBP/C operator