Multi-resolution image segmentation method based on local variance

An efficient image segmentation method for complexity image is presented, which is developed based on the evolution of local variance in multi-resolution images for different objects. The proposed method first introduces a multi-resolution model in which pixels between different resolutions are linked by computing parent-child distance. Then the evolution of local variance across resolution is studied. By counting the residue of local variance of image at different resolutions, the distinction between object and background is amplified. After a blurring procedure, the residue is segmented with Ostu threshold method. Finally, the segmentation result of original image is obtained through projection with parent-child link. The experimental results show the excellent performance of the new method.