A HVS-driven image segmentation framework using a local segmentation performance measure

This paper presents a novel framework for image segmentation. In this framework, image segmentation is considered to be a detection problem, and a “soft” segmentation objective function, in terms of the detection performance measure in local regions, is employed to guide the segmentation procedure. The human visual system information is also incorporated into the segmentation procedure to improve the efficiency of the framework by introducing a contrast sensitivity function-weighting operation in the wavelet domain. Encouraging experimental results are obtained when the algorithm is applied to real-world image data.