Objective And Subjective Assessment Of Perceptual Factors In HDR Content Processing

High Dynamic Range (HDR) images could capture much higher contrasts and moredetails as compared to the traditional8-bit low dynamic range (LDR) signals. It isexpected to be widely used in digital cinema, remote sensing and next generation broadcast,owing to its high quality and its descriptions of sceness in naturalness. However, retrievingHDR images would face difficulties of inferring the tone mapping equation and the many toone mapping even in a unique equation. Visual distortions resulted from different tonemapping methods should be investigated, while the pending research of objective evalutiondemands more accuratr solutions. The main work is as follows:1. Tone mapping and inverse tone mapping operator can cause distortion, In this paper,we propose a scale controllable logarithmic feedback tone mapping method. We constructed acontrollable scale logarithmic equation based on Weber-Fechner law, and then we use the tonemapping model that is optimized to minimize the error between the HDRI and its tonemapped version. Our method outperforms the conventional method in senses of the MSE ofHDR images. On the compression ratio from0to0.1our method makes MSE valuedecreased by approximately10%. The quality of LDR images get from our method is alsobetter than the old one gets.2. In accordance with the International Telecommunication Union (ITU-T P.910)standard, we did a subjective image quality testing for the HDR images. We analyzed theimpact of TMOs on the visual quality using a total of6264ratings obtained from thesubjective tests. When the compression ratio is high, tone mapping methods are the mainfactors that affect the quality of decompressed image, so it is better to set the HDR imagecompression rate between0.3to0.4. When the dynamic range span is large, use different tonemapping methods can cause bigger difference for the generate HDR images. Different tonemapping operators indeed lead to quite different perceptual quality in the decompressed HDRimages. Ashikhmin tone mapping method gets the best results. Further, we carried outstatistical analysis and found that statistical differences exist between the visual qualityratings of the images processed by different tone mapping operators. 3. HDR-VDP-2is a visibility prediction metric which also employs a pooling function tocompute an overall quality score. This paper focuses on the pooling aspect in HDR-VDP-2used the HDR database for improving the prediction performance of HDR-VDP-2byfinding better pooling parameters. This was done by minimizing the error between thelogistically transformed predicted values and the subjective ratings. We use a database ofHDR images (with their corresponding subjective ratings) to improve the prediction accuracyof HDR-VDP-2. The use of HDR images for parameter optimization lead to an overall betterperformance.