A Fast Texture Feature Extraction Method for Region- based Image Segmentation

Region-based image segmentation is one popular approach to image segmentation of generic images. In these methods, an image is partitioned into connected regions by grouping neighboring pixels of similar features, and adjacent regions are then merged with respect to the similarities between the features in these regions. To achieve fine-grain segmentation at the pixel level, we must be able to define features on a per-pixel basis. This is straightforward for color information, but not for texture. Typically texture feature extraction is very computationally intensive for individual pixels. In this paper, we propose a novel fast texture feature extraction method which takes advantage of the similarities between the neighboring pixels. The experiments demonstrate that our method can greatly increase the extraction speed while keeping the distortion within a reasonable range.