Research Of Remote Sensing Image Quality Assessment Method Based On Edge Feature

Remote sensing image is affected by platform shock of imaging system and the moving of the target and sensor in the process of acquirement. It leads to the descending of image. The quality of the remote sensing image affected the later image processing and application directly. Therefore, the study on remote sensing image quality assessment is important.This thesis mainly studies the remote sensing image quality assessment method based on edge feature which faces the object recognition. This method chooses several feature descriptors by the human vision characteristics, builds image quality assessment factors by feature matching, determines the composite index by weighted sum and then accomplishes the assessment of remote sensing image quality. In addition, this thesis combines the method with neural network to build remote sensing image quality assessment model when the image numbers is sufficient so that improve the efficiency of image quality assessment.The content and innovation points of this thesis can be concluded as follows:Firstly, study the physical structure and the vision perception characteristics of human vision system, conclude the relationship of edge feature and human vision system. It demonstrates the rationality of evaluation remote sensing image quality based on the edge feature in the theory.Secondly, study the image preprocessing method and edge detection method. Analysis the characteristics of remote sensing image, choose proper image enhancement method. Research the typical edge feature detection means and figure out their shortcomings, and then put forward better edge detection method. This method has good accuracy of the detection result, better noise immunity and less false edges. The experiment result proves the advantages of this new method.Thirdly, study feature description method of remote sensing images. According to the background of this thesis, choose shape context, edge feature ratio, weighted gradient and the ratio of length and width of line pairs. This thesis studies deeply the remote sensing image quality assessment method based on square gradient and point acuteness and figure out the first method has bad noise immunity and the second method has better noise immunity and better sensitivity.Lastly, study the remote sensing image quality assessment method based on edge feature. This method builds image quality assessment factors by feature matching, determines the composite index by weighted sum and then accomplishes the assessment of remote sensing image quality. In addition, this thesis combines the method with neural network to build remote sensing image quality assessment model when the image numbers is sufficient so that improve the efficiency of image quality assessment. The experiment method proves that it has good noise immunity and high sensitivity. It can be assess the quality of image which contains single distortion factor and several distortion factors. In addition, the experiments result of second remote sensing image quality assessment method proves that it can assess image quality intelligently and improve the efficiency of image quality assessment.