Image Quality Assessment Based On Human Visual System

The popularization of personal computer and Internet prompts the communication and sharing of digital images. However, during the procedure of image acquisition, synthesis, enhancement, compression and transmission, the image signal will be altered, which will obviously impact on the visual quality of the images. There are significant implications on researching the methods of image quality assessment(IQA), and developing effective IQA metrics. It can be applied to the evaluation of image quality, the monitoring of image quality, the optimization of image processing algorithms, the parameter setting of imaging system and etc.Despite of the varieties of image quality methods, the comprehension and application of the human visual characteristics are irregularity. Additionally, few efforts were made on the medical image quality assessment and the result of which cannot meet people’s expectation well. Efforts have been made as such in this dissertation.(1) An overview and analysis have been made based on the properties of human visualsystem. Human visual system usually has different sensitivity on different structure andit will pay more attention on the boundaries. Based on that, many attempts have beenmade to propose an effective model of human visual characteristics.(2) In the context of the detailed overview and analysis of existing IQA researches, thestructural saliency model has been proposed, and then which is combined with thephase congruency and gradient magnitude to get a comprehensive IQA metric. Based on3 mostly used public image quality assessment benchmark databases, extensiveexperiments have been made to prove that the metric can be used to correlate well withthe subjective evaluation.(3) This dissertation also established the medical image quality benchmark database formedical image quality assessment research. Wherein, 300 images have been evaluatedby 4 radiologists. Furthermore, the IQA metric based on structural saliency has beentested on this database, more satisfactory results can be achieved which outstands othertraditional methods.Based on the characteristic of human visual structural saliency, an effective IQA metric is proposed, which can reach satisfying result. It has significance on improving the reliability of objective IQA work. This database can be further used to test new metrics and provide uniform platform for the comparison between different metrics.