(A) study on image quality assessment methods using image structure and contrast characteristics

Abstract

iii) We propose one more IQA method, called DCT-based Quality Deg-radation Metric (DCT-QM), that is especially effective for perceptual video coding (PVC) applications. The DCT-QM is derived from a low-level psycho-visual theory and is finally formulated as a weighted mean $L^2$ norm for DCT-coefficients in the DCT domain. Our DCT-QM is very easy to be applied in PVC schemes without additional complex computations (e.g., DCT operations for original image blocks to be encoded) and inherits the three desirable mathematical properties for image quality optimization problems, that is, a valid metricability, differentiability and convexity for DCT coefficient values in the DCT domain. Our DCT-QM utilizes spatial CSF characteristics in the DCT-domain but do not adopt the SCI feature used in SC-QI (SC-DM). This is because DCT-QM is aimed to directly and effectively be applied in PVC schemes. So, DCT-QM can be viewed as a simplified version of SC-QI (SC-DM). Extensive experimental results show that both SC-QI and SC-DM outperform the state-of-the-art IQA methods with much lower computation complexities. The proposed DCT-QM also shows competitive per-formance in estimating perceived visual quality compared to the state-of-the-art IQA methods. In addition, a fast DCT-QM version is proposed which lowers computational complexity with significantly faster running speed and competitive prediction performance compared to the original DCT-QM. In consequence, our SC-QI, SC-DM and DCT-QM can have great merit to be applied into such image/video coding and processing applications.

Computational (or objective) image quality assessment (IQA) is one of the most fundamental yet challenging problems in image processing and computer vision fields. Recently, there has been increasing in-terest in developing elaborate objective IQA methods. Objective IQA methods can effectively replace subjec-tive image quality assessment experiments which are often time-consuming, expensive and difficult to be implemented in systems. Also, they can be effectively applied in image quality optimization problems (e.g., optimal image restoration) which aim at maximizing visual quality metric values in objective functions. In this regard, there has been much effort to develop objective IQA methods having both high correlations with visual quality perception and desirable mathematical properties to be applied in image quality optimization problems. In this study, we proposed three IQA methods that have different application purposes: i) We propose a new IQA method, called Structural Contrast-Quality Index (SC-QI), by adopting a structural contrast index (SCI) which can adaptively characterize local visual quality perception depending on different image charac-teristics and structural distortion types. In addition to SCI, we devise some other perceptually important fea-tures for our SC-QI that can effectively reflect the characteristics of human visual system (HVS) for spatial contrast sensitivity function (CSF) and chrominance component variation. The proposed SC-QI shows the best correlations with visual quality perception compared to the state-of-the-art IQA methods

ii) In addition, we extend SC-QI to structural contrast distortion metric (SC-DM) which inherits desirable mathematical properties of valid distance metricability and quasi-convexity. So, it can effectively be used as a distance metric for image quality optimization