The Combination Of Subjective And Objective Of Remote Sensing Image Quality Evaluation Method Research

During collection, processing, transmission and recording of remote sensing images, image distortion and degradation is emerging generally. The reason includes the instability of the imaging system, the inappropriate of processing methods, and unreliable transmission mediums/recording equipments. Therefore, the image quality assessment method is an integral part of the remote sensing system, which can analyze the effect of image acquisition/processing and the feedback of image transmission/recording quality. Moreover, it helps to set the remote sensor parameters adaptively. According to the difference of evaluated objects, the image quality assessment method includes subjective and objective evaluation methods.Our Specific work on subjective and objective image quality evaluation methods includes the following three aspects.First, we establish an expert grading library of remote sensing images. We collect remote sensing images as the ground truth and utilize them to generate distorted images of the simulation. This paper employs the state-of-the-arts of subjective quality assessment methods and sets a special laboratory environment to score those distorted images. We rely on the generated expert grading library, in the following research and implementation of objective quality assessment methods.Second, this paper introduces Human Visual System (HVS) into structural similarity (SSIM) model for a full-reference quality assessment algorithm of remote sensing images.1.We use the brightness contrast, texture complexity and the spatial positions to generate the original visual perception figures. Then distorted perception figures are calculated.2.According to the distorted perception figures, we locate the distorted regions. Further, since significant visual features and serious distorted regions will cause a shift in focus of visual attention, we can estimate new visual perceptual images.3.Via weighting the block similarity of the original and generated visual perceptual images, we achieve the final objective quality assessment results.Third, we fusion HVS and the contrast-based/fuzzy-based models, where two no reference quality assessment algorithms for remote sensing images are accomplished. In the contrast-based model, the paper introduces two weighting factors:the visual region of interest and non-vision region of interest via analyzing the feature of interest of the human visual system. This leads to the construction of the contrast-based model with HVS. In the fuzzy-based model, we consider the mask characteristics of HVS and introduce the brightness mask and the space complexity models, where the fuzzy-based model is generated.