Texture-based Image Segmentation Method Study

Nowadays Texture image segmentation is a hot research topic, because it is important for Image Processing, Pattern Recognition and Computer Vision. There is a lot of application about it, such as cancer cell recognition in medical diagnosis, distinguish military objectives and civilian targets from remote sensing image, etc.Texture image segmentation means that an image which is composed by different textures is divided into several regions which have the same or similar texture. Texture image segmentation contains two parts: feature extraction and region consistency segmentation algorithm. The result of Texture image segmentation is a set which contains labels that are used to label every pixel.Before this paper, the following works have been done: investigated main texture feature extraction methods by now, such as Spatial autocorrelation, Fourier power spectrum, Co-occurrence matrix, Neighborhood characteristics based on statistical texture analysis, Gray level difference statistics, stroke length statistics, fractal dimension texture description method, Tamura texture features method; investigated the processing mechanism of the human visual system; investigated "Filter -> Full-wave rectifier -> filter" (FRF) model of texture segmentation based on HVS; written MATLAB program to achieve texture segmentation algorithm based on FRF model; tested and debugged the written MATLAB program with Brodatz's texture images library, to verify that the texture segmentation algorithm based on FRF model is correct and how is its resolution. Implemented the above algorithms with VC and ported to project platform.