Fast segmentation of texture image regions based on hill-climbing

In this paper, we present a fast method for segmenting texture images. Since the texture features of images are generally present at various scales, a multiscale decomposition method is necessary to analyze the image effectively. Thus, this paper utilizes Gabor filters to extract texture features of images. Then, we use a simple and fast hill-climbing algorithm to detect coherent regions of an image. Our hill-climbing algorithm detects the peaks (local maxima) that represent clusters in the global texture histogram of an image. We utilize the histogram bins rather than the pixels themselves to find these peaks; thus, our algorithm can find the peaks efficiently.