Multiple Granularities Self-Modified Method Based on MRF for Image Segmentation

In this paper, a multiple granularities self-modified method based on MRF (MGSM-MRF) is proposed for image segmentation. Current segmentation methods are usually focused on a single pixel-based granularity of an image. But the multiple granularities property of an image is limited to the methods using a single granularity. The proposed MGSM-MRF considers the multiple granularities property for image segmentation by incorporating MRF. The advantages of MGSM-MRF include the multiple granularities description for an image and a self-modified method for segmentation by different granularities. Experiments prove a high accuracy and efficiency of MGSM-MRF compared with single pixel-based granularity MRF model.