Research On Subjective And Objective Quality Assessment Of Bit-depth Enhanced Image

With the development of hardware equipment,people are pursuing higher and higher quality of images.The improvement of image quality has become the focus of people’s attention,and image enhancement technology in various subdivided fields has also emerged.The bit-depth enhancement technology refines the color accuracy by reconstructing the least significant bits of the image,and provides a more vivid and delicate visual experience.However,due to it is a fledgling branch,there are some different degrees of artifacts in high-bit images reconstructed by different algorithms,such as color offset,false contour,or excessively blurred texture,which seriously affect the subjective visual perception.For this kind of special distortion,the existing general quality evaluation algorithms are not suitable,an effective and accurate special objective quality assessment algorithm is urgently needed to evaluate and guide the bitdepth enhancement algorithm.The design of bit-depth enhanced image objective quality assessment algorithm needs the support of enhanced image subjective assessment database Therefore,based on the research and analysis of the bit-depth enhancement algorithm and its distortion characteristics,the main research contents of this paper are as follows:1)Aiming at the quality assessment of bit-depth enhancement,a special database for bit-depth enhanced image quality assessment is constructed.Collect suitable highquality images with rich content and different styles,discard least significant bits as bitenhanced materials,and use eight representative bit-depth enhancement algorithms to generate reconstructed high-bit images with 4 to 8 and 4 to 16 two kinds of bit-enhanced spans,set up experimental links,design and implement subjective quality evaluation experiments,obtain subjective opinion scores of images,and construct image quality evaluation database.On this basis,the subjective perception of bit-depth enhanced image distortion is analyzed.2)A full-reference quality assessment algorithm for bit-depth enhanced images is designed.Based on the analysis of the causes and visual characteristics of the bit-depth enhanced image distortion,the difference and similarity between the distorted image and the reference image are detected from the color,structure,quantization and other aspects.The image visual quality is mainly evaluated by using image information such as color,gradient,histogram and so on.Experiments show that,the proposed algorithm has better quality evaluation performance compared with the existing full-reference quality assessment algorithms.In addition,when the bit-depth enhancement span is known,the algorithm can be degraded to a reduced reference quality assessment algorithm by replacing the reference image with a low-bit image,and it also has good performance.