A New Image Quality Approach Based on Decision Fusion英文资料.pdfVIP

Fifth International Conference on Fuzzy Systems and Knowledge Discovery A New Image Quality Approach Based on Decision Fusion Mingna Liu and Xin Yang Institute of Image Processing and Pattern Recognition, Shanghai Jiaotong University, Shanghai, 200240, P. R. China {mingnal, yangxin}@ Abstract the redundant information among them to certain degree[8]. Some previous metrics dealing with multi-feature in- Image quality evaluation is becoming essential in many tegrations handle similar features or specific applications, image processing problems. This paper proposes a new im- which limit their capability in some different distortion age quality evaluation approach based on decision fusion conditions[1, 3, 9]. In this paper a supervised multi- method of canonical correlation analysis (CCA). By explor- feature (SMF) image quality evaluation metric toward solv- ing several diverse visual models, we constructed a com- ing the problem of complicated image distortion is pre- prehensive quality metric which can deal with complicated sented. Some the state-of-the-art image quality metrics ex- image distortion problem with increasing accuracy and ro- ploring different visual models are merged into a compre- bustness. Validation by comparing the proposed metric hensive one by the decision fusion method of canonical cor- against other image quality metrics (IQMs) demonstrates