Quality Evaluation Methods Research And Software Design For Remote Sensing Images

Remote sensing is becoming more and more popular in agriculture,water conservancy,military and so on.The quality of remote sensing image directly affect the accuracy of monitoring,forecast and reconnaissance in various areas.Besides,it can also reflect the performance of the remote sensing imaging system.Remote sensing image quality evaluation plays a very important role in remote sensing application.This paper mainly do the research in several following aspects:1)Factors affecting the quality of remote sensing image.Imaging principle and imaging system are discussed,and several factors resulting the image degradation mainly focus on blur,noise and cloud cover.2)A single distortion evaluation of remote sensing image.As for blur,the evaluation algorithms based on image histogram concentration and based on gray-gradient co-occurrence matrix are deeply analyzed.As for noise,the evaluation methods based on local standard deviation and based on the statistical feature of wavelet domain are deeply studied.As for cloud detection,the algorithms based on FCM and based on multi-threshold are deeply analyzed.Then the simulation experiences are done to compare the advantages and disadvantages of these algorithms above,and the suited situations are decided.3)Multi-distortion evaluation of remote sensing image.When noise and blur are mix up in an image,the single distortion evaluation algorithm cannot perform well,so the evaluation algorithm suitable for multi-distortion is proposed.It extracts the characteristics of natural scene statistics,and calculates the main parameters of probability density distribution function as a characteristics value,then uses the support vector regression to get the final noise value.4)Remote sensing image quality evaluation software design.The overall architecture and process are proposed,and the function of blur evaluation module,noise evaluation module,cloud detection module,comprehensive judgment module are introduced.Besides,it specifically introduces the main algorithms,then shows the computational results and how to operate on the software interface.