Image Quality Evaluation Method Based On Dual Scale Edge Structure Similarity Of ROI

Image quality assessment plays an essential role in the field of medicine, image compression, processing and communication. The health information of human body can be reflected in the form of images with the development and application of the advanced imaging technologies, such as Positron Emission Tomography and Computed Tomography, etc. Medical image is able to show the structure of body intuitively and accurately, the quality of medical image will seriously affect the diagnostic result of doctors.Traditional image quality assessment methods based on error sensitivity are applied commonly to clinical practices, which is simple to image calculation and has clear physical meaning, but has no good visual feelings of human subjective. The philosophy of SSIM models the functionality of the overall HVS top-down, the SSIM method assumes that, the main function of the HVS is to extract structural information from viewing fields.The SSIM algorithm can better reflect the characteristics of human vision, it is simplified and efficient. However, the assessment result of SSIM has a deviation between blurred, contaminated image and human subjective’s feeling. A dual-scale image quality assessment method was proposed to solve this weakness. In this algorithm, the edge information is appreciated sufficiently in the structure information. The algorithm has effectively solved the shortcomings of SSIM, but has not taken into account the importance of the region of interest in the quality of the image, and it is not consistent with the subjective human feelings when evaluating images.In this paper, we study the algorithm of SSIM and MDESSIM in-depth, and summarize their laws and defects, and make improvements based on SSIM.A strategy. ROIMDE is proposed based on region of interest (ROI) and dual-scale edge structure similarity (SSIM). We apply the Itti model to extract the region of interest and region of non-interest by analyzing the currently ROI extraction algorithm. The quality assessment of the ROIMDE image is regarded as a weighted combination of ROI and NON-ROI, the dual-scale edge structure similarity is used in ROI, and the classical structure similarity is applied to NON-ROI, and a greater weight is given in the region of interest in order to emphasize the importance of the ROI, and we also discuss the value of weights set in the experiment. Some experiments are conducted by three common image databases with the proposed algorithm, and the results show that, the proposed method is more consistent with human subjective visual evaluation by considering the importance of the Region of Interest.