Image Quality Assessment Based On Hvs

Images can provide more affluently and lively information to human than other kinds of media, such as sound and texts. What is called 'ear listens for empty, the eye sees for solid'. Now another word is more popular on the network which is 'picture is truth'. These mirror the importance of images for life. But images are subject to various types of distortions during acquisition, compression, transmission, processing and reproduction. If the distortion is little, it will make the image looks inaesthetic. If the distortion is serious, it can cause twisted understanding or even we can't understand it at all. The purpose of the image quality assessment is looking for an ideal mathematical model to evaluate the quality of the image, making the evaluation result agree with the result of subjective image quality assessment. The significance of the image directly promotes the research on image quality assessment, and achieved good results.At the beginning of the paper, we introduced the subjective image quality assessment methods and several popular image databases. After the study of the theory of image quality assessment and classic objective image quality assessment methods, we put forward a new image quality assessment method, which is based on the multi-scale geometric analysis and phase congruency. As is known to all that human visual system is very important to image quality assessment method, because a person observes and understands an image by the visual system. If one method can consider all the human visual characteristics, it will get a perfect result, while the human visual system is very complicated, we still know it very little until now. What we should do is to continue to study the characteristics of the human visual system, and to look for the best model by using the characteristics of image quality assessment that we already know. In this paper, a new objective method feature-similarity and multi-scale geometric analysis (FMGA) is presented. It assesses image quality by using different characteristics of human visual system (HVS). Human understands an image mainly according its low-level features, such as edges. Specifically, the phase congruency (PC) is the best way to describe it. Considering PC is contrast invariant while the contrast information does affect HVS'perception of image quality, the image Gradient magnitude (GM) is employed as the secondary feature in FMGA. Considering HVS is multi-channel, the multi-scale geometric analysis(MGA) is employed as the tertiary feature in FMGA. PC, GM and MGA play complementary roles in characterizing the image quality.With LIVE(Laboratory for Image & Video Engineering) and TID2008(TAMPERE IMAGE DATABASE 2008 TID2008) image database, we compared the proposed method feature similarity and multi-scale geometric analysis(FMGA) index with feature similarity(FSIM) index and multi-scale geometric analysis(MGA) index. The results demonstrate that FMGA has a higher consistency with subjective assessment than the others.