Based On Singular Value Decomposition Of The Objective Image Quality Evaluation Characteristics Research

In the process of image acquisition, compression, storage, transmission and reproduction, digital images will often suffer different levels and different types of distortion. These distortions will influence the image quality. It is very important to be able to assess the quality of the image so that it can maintain, control and possibly enhance the quality of the image. Hence, a reliable image quality metric (IQM) is very crucial in many image processing fields.In order to solve this problem, we will explore the feasibility of singular values decomposition (SVD) in developing a novel metric that can express the quality of distorted images. The novel objective metric uses the singular vectors and singular values as the features to evaluate the quality of the distorted image from the view of energy and structure. With the help of quaternions, this metric will be extended to color image quality assessment. As to correlate well with human visual system, attention selection will be introduced to the image quality assessment metric by using image saliency map with a little increment of computation. The saliency-based metric shows a great improvement in the consistency with the human visual system. The main contributions of this thesis are as follows:1. Propose a novel yet effective full-parameter image quality assessment metric by employing SVD. The novel objective metric uses the singular vectors and singular values as the features to evaluate the quality of the distorted image from the view of energy and structure. Many tests are conducted for evaluating the performance. Compared with traditional algorithms, the proposed method shows a great improvement in the consistency with human visual system (HVS).2. Present a new metric to evaluate the quality of color image from the view of structure distortion, energy distortion and color distortion. This metric uses polar form of quaternion to represent the color image. Theoretical analysis shows that the norm of the quaternion contains the luminance information and the eigenaxis covers the color information. Therefore, the singular vectors and singular values of the quaternion’s norm are used to obtain the structure and energy distortion of the image. The eigenaxis is used to calculate the color distortion. Compared with traditional metrics, the proposed method can effectively evaluate the energy distortion, structure distortion and color distortion of the color images.3. Put forward a novel saliency-based image quality assessment metric. We extract the saliency map from the reference image by PFT and PQFT. Then the saliency map is used to weight the metric in order to improve the performance of IQM. Compared with existing methods, saliency-based metric are more consistent with the subjective evaluation.