No-Reference Remote Sensing Image Quality Assessment Using Comprehensive Assessment Algorithm

In the imaging process of satellite remote sensing images, due to the effect of various distortion factors, leading to the obtained remote sensing image data existing quality degradation and resolution is reduced, further affecting the subsequent processing and application of remote sensing image data. In this paper, through the analysis of remote sensing images, aiming at two kinds of circumstances, stripe noise and no-stripe noise remote sensing image data, respectively. Remote sensing image quality comprehensive evaluation model is established for the image quality evaluation,la order to evaluate the different distortion types of remote sensing image quality, it is very necessary to research for a comprehensive evaluation index aiming at a.wide variety of quality degradation factors. First of all, MTF of remote sensing image is calculated based on the improving knife edge method, and a no-reference remote sensing image quality evaluation method based on MTF is studied. In view of the limitations of the traditional edge, combining an improved Canny operator with Hough transform operator, the improving edge method contains no artificial factors, and extracts suitable edge self-adaptively from the remote sensing image. Compared with the traditional edge method, the improved Canny operator can detect the image edge and feature point, and it is a single response and gets high precision of MTF curve, making remote sensing image quality evaluation based on MTF more accurate. Secondly, due to different degrees of noise and blur exists in remote sensing image, combining HVS and SSIM, the noise and blur assessment parameter VSSIM can be obtained. By the human visual system perception of blur and noise features, it is need to measure the variation of visual perception information before and after image filtering in the edge area, as well as the variation of visual perception information before and after adding noise in the smoothing area. SSIM is used as the measurement of this kind of variation, and Phase Congruency is used to distinguish the different areas of the image. SSIM values are calculated in the edge area, as well as the SSIM values in the smooth area. The greater the SSIM values, the smaller degree of quality change between corresponding images area, the opposite is larger. VSSIM value is taken as quality evaluation result according to the quality variation in different areas and the evaluation index VSSIM conforms to the human visual system without reference images, which can reflect the visual perception of image quality well. Aiming at remote sensing image including stripe noise, which has certain periodicity, directivity and zonal distribution, Phase Congruency can locate the stripe noise of remote sensing images and is used to acquire the stripe noise parameter NIs without the original reference images, and the assessment method is not affected by changes of the image contrast. In the condition of the brightness changes reliable image visual perception characteristics is obtained, the assessment method can effectively evaluate stripe noise of remote sensing images. Finally, considering various distortions, including the quality of imaging system, common noise, blur and stripe noise factors, according to all the single factor evaluation characteristic value of remote sensing images, the mathematical model of comprehensive evaluation factors is established based on the Choquet fuzzy integral, and fuzzy integral value is derived as the comprehensive evaluation parameters at last, aiming at two kinds of circumstances, stripe noise and no-stripe noise remote sensing image data, respectively.The experimental results show that the comprehensive evaluation model can effectively evaluate the quality of remote sensing images. The original reference images are not required, and the evaluation result is consistent with the human visual system, has certain validity and extensive applicability.