Image Quality Evaluation Model Based on Local Features and Segmentation

The perceived image distortion of any image is strongly depend on the local features, such as edge, flat and texture. A new objective no-reference (NR) image quality evaluation model based on local features and segmentation for JPEG coded image is presented in our previous paper, which is easy to calculate and applicable to various image processing applications. But the algorithmic thresholds investigation of the segmentation algorithm were not sufficient in the paper. Therefore in this paper, we want to investigate the suitable threshold values of our segmentation algorithm both for training and test images by the optimization method. Our experiments on various image distortion types indicate that its performs significantly better than the conventional model.