PERCEPTUAL POOLING STRATEGIES FOR IMAGE SEGMENTATION QUALITY EVALUATION

Image segmentation is an essential step for many computer vision tasks. In this paper, we propose two pooling strategies to evaluate the image segmentation quality. Based on the hypotheses that correlate with the human perception of segmentation quality, we explore to assign perceptual meaningful weights to the quality map. To the best of our knowledge, this is the first work that adopts perceptual pooling strategies in the quantitative segmentation evaluation. Extensive experiments are conducted on the subjective evaluation benchmark and BSDS500, which indicate that the proposed strategies can improve the performance of evaluation measures and produce a more perceptually meaningful judgment on the segmentation quality.