Research Of Color Image Quality Assessment Based On Expanded Uniform Color Difference Space

In modern society, image information has entered our household’s daily life such as in digital image processing system of digital TV, video conference and video phone. Due to the faultiness of imaging system, processing method, transmission media and storage equipment, image will be inevitably distorted and deformed. Thus image quality assessment becomes more and more important and attracts more and more researchers. Presently the main research direction in this field is objective image quality assessment. The goal is to make the objective quality assessment model accurately reflect the subjective quality of human vision. The focus is to reduce deviation between the subjective and the objective quality assessment results. Color vision system is more complex compared with monochrome image, thus it is more involved to establish an objective color image quality assessment model. For this challenging research topic, every new result can be regarded as an important progress and innovation of image quality assessment theory.Normally, color image quality assessment normally maps three primary colors into three elements space or uniform color difference space and then uses monochrome image quality assessment to process respectively. However, this kind of methods based on the three elements space lacks the support of uniform color difference space. Similarly, the methods based on the uniform color difference space need complement of other effective evaluation model. In this dissertation, we proposed an objective color image quality assessment based on extended uniform color difference space by combining uniform color difference space model, color space structure similarity decomposition model and HVS (Human Visual System) model. This method covers the physical characteristic and the corresponding mathematical model of basic visual experiment. In addition, it can explain the uncertainty of independent assessment model result and improve the effectiveness and stability of assessment model. The content and innovation are listed as following:1. In contrast with previous methods lacking direct comparability with human subjective feeling, this dissertation firstly presents a uniform color space model which can achieve a linear relationship between changes of brightness, hue, saturation of subjective perception. The essence of uniform color difference space model is an expansion of the brightness, hue and saturation point for Web-Fechner law. In order to investigate the subjective perception of luminance, hue and saturation, this dissertation also designs a set of test image signals of brightness, hue and saturation utilizing the consistent characteristic of CIEDE2000color difference formula and the subjective perception.2. As visual structure similarity component in color error image can cause uncertain phenomenon in normal color image evaluation model, this dissertation presents a color space structure similarity decomposition model by extending the HVS model in shapes. This uniform color difference space structure similarity decomposition model combines color space structure similarity decomposition model and uniform color difference model so that it can be used to investigate the subjective perception of structure similarity decomposition model in uniform color difference space. This method can be regarded as a extension of traditional image quality assessment methods from the aspects of structure and color.3. HVS’s multichannel decomposition model mainly investigates the influence of object space size on the subjective vision. This dissertation presents a convenient multichannel decomposition model of DCT gauss quantization model. This method can be used to investigate the subjective perception of object space size in uniform color difference space by combining multichannel decomposition model of DCT gauss quantization model and uniform color difference model.4. HVS’s visual masking space model reflects visual perception phenomenon of "existing an incentive can lead to change of another incentive threshold". Different from the previous masking models mainly focus on gray image, this dissertation presents a color space comprehensive visual masking model. In addition, a uniform color difference space visual masking model is proposed by combining color space comprehensive visual masking model and uniform color difference model. Therefore, the subjective perception of visual masking model in uniform color difference space can be investigated.5. This dissertation proposes a uniform color difference space extension model which can assess color image quality more effective, stable and comprehensive by combing uniform color difference model, color space structure similarity decomposition model and HVS mode. This method can be applied in various conditions and can make assessment results more coincide with subjective perception.