Texture Image Segmentation Using Gabor Filter and Anisotropic Diffusion Filter

Abstract

Image segmentation is very important task in many imageanalysis or computer vision applications. In this paper a textureimage segmentation method using Gabor filter, anisotropic filter,and k-mean clustering algorithm was proposed. The Gabor filterwas used as a multi-channels filter to analyze the texture in theimage. The extraction and enhancement of the texture featuresobtained using anisotropic diffusion filter. Then the k-meanalgorithm used to cluster pixels into number of clustersrepresenting the texture regions. The quality of segmentationusing this method was evaluated using Ultimate MeasurementAccuracy (UMA) metric. The experiments show that theperformance of this method is effective, accurate and gives betterresults as compared with the Seo method from the view of qualityof segmentation, the number of run times, the execution time andthe capability of separating a large number of textures, and ofsegmented real images, random mosaics texture images, area ofroofs and ground images, and to distinguish objects frombackground.