Study On Haze Removal Method And Parallel Implementation For HJ-1 Satellite CCD Image

HJ-satellite is one of the satellites of"Environmental Monitoring and Disaster Forecasting satellite constellation",which is to dynamic monitoring the environment and disasters of our country, neighboring countries and regions. Now, with the effectiveness of satellite application and the increase of the number of HJ-1 CCD images, researching the processing method is imminent. The research about haze removal from remote sensing image is the basis of remote sensing applications, and it is an important part of the remote sensing image processing. This paper aim at the data characteristics of HJ-1 satellite CCD images, study the haze removal method being the same with HJ-1 satellite CCD images, the main contents are as follows:(a)HJ-1 satellite CCD images haze removal methods. Through the analysis of characteristics of HJ-1 satellite CCD image data, based on"BSHTI"(background suppressed haze thickness index) cloud detection method and the"virtual cloud point"haze removal method, this paper improved and perfected the method. And the advantages and limitations are summarized by comparing this algorithm with original algorithm through experimental.(b) HJ-1 satellite CCD images thick cloud removal methods. First, research on the method of radiation correction between the dual-phase images, and put forward new ideas of the selection of Pseudo-Invariant Features in"Pseudo-Invariant Features"(PIFs) method, which is match the feature points as the"Pseudo-Invariant Features". Then, replace the cloud image based on the two-phase images, and lead into the idea of the eclosion borders, and appropriate adjustments were carried out. So the problem of the radiance difference between the area of cloud and the original image is resolved.(c) Design and realize parallelization for cloud removal algorithm of remote sensing image. In order to increase the speed and efficiency of the cloud removal methods, this thesis paralyzed the haze removal method based on MPI (Message Passing Interface) parallel processing technology. First, analyze the parallelization of the two methods, then design and implement parallel programs of these two algorithms, and last the performance evaluation of the parallel program is given.