摘要
The research is motivated by the conclusion that single metric can't give the best assessment in all situations. In this paper, a new method for objective no-reference (NR) image quality assessment (IQA) with multi-metric combination (MMC) is presented, and has four characteristics: 1) It is a NR assessment applicable for arbitrary multi-focus images; 2) The metric-selection based on segmentation consistency is effective to process various multi-focus images. 3) The linear and nonlinear conversion functions are constructed respectively to transform each pair of selected metrics into a set of basic probability assignment (BPA). 4) D-S evidence theory is introduced to exploit the advantage of performance evaluation and explore a resolution to the conflicting circumstance. Experiments conducted with five sets of multi-focus images confirm the effectiveness and robustness of our metric in contrast with single metric.
- 出版日期2017
- 单位广东工业大学