Base On Improved Algorithm Of POCS-MAP For Remote Sensing Image Super Resolution Reconstruction

Remote Sensing Technology today has made rapid development; it can achieveobserving the same area by multi-platform, multi-sensor and all-weather. So it will geta lot of repeat remote sensing images. Because we do not use the information of a lotof the remote sensing images. There are two reasons: Firstly, by the hardwarelimitations of the sensor, ground information we get is due to the sampling, so we geta low-resolution image of the actual surface. It is unable to meet the requirements ofapplication; secondly, due to the remote sensor platform at high altitude, the imagingsystem has been impacted by atmospheric disturbance, light effects. Remote sensingimages produced a lot of motion blur and random noise cannot be avoided. In order touse of a large number of remote sensing images effectively, meeting the application ofremote sensing images, many people are through a variety of ways to improve thequality of remote sensing images.According to the above-mentioned factors affect image quality, improve remotesensing image resolution are generally two ways: one is to improve the hardware level,that is to say, to improve the resolution of the sensor, so the pixel size can be reduced,increasing the number of pixels per unit area, but this will reduce the signal-to-noiseratio of the image, and can also increase the imaging chip size, but this will increasethe capacitance, making the charge-transfer speed drops. In addition, a precisionsensor costly and limitations of the manufacturing process makes this methodencounters a bottleneck. Another method is written by software algorithm to achievehigh-resolution remote sensing image, this method does not require expensiveequipment and still take advantage of the wealth of low-resolution images. The superresolution of image is one of the effective methods.Remote sensing image super-resolution reconstruction is the different angles ofthe same area, different observation time even different sensor degradation of theoptical system of the low-resolution remote sensing images, these low-resolutionimages as input, after the one closest to the original high-resolution imagereconstruction algorithm output process.The process of rebuilding, restoring lose high-frequency information, the information outside of the reconstruction of thecutoff frequency of the imaging system. The thesis is based on this idea; the two setsof experiments are based on multi-frame images of the same area to rebuild theregion’s high-resolution images.The basic principle is that using of the image sequence of the low-resolution’sinput information and prior knowledge will be add to restore the reconstruction of acomprehensive infinitely close to the high-resolution image of the original image.This thesis introduces the imaging mechanism of the optical remote sensingimage firstly, and then theoretically proves the feasibility of super-resolution remotesensing image and reconstruction capabilities. Subsequently, the super-resolutionrelates to all aspects of make improvements, in particular made improvements on thealgorithm itself, made in the traditional POCS super-resolution algorithm based on thePOCS algorithm, and thus together with the improved MAP algorithm. Improvementideas generally have the following aspects:Firstly, the POCS algorithm for super-resolution reconstruction has someproblems of the heavy dependence of the initial value and the edge oscillation, thefractal interpolation is used to replace the general bilinear interpolation in this thesis;secondly, select the appropriate threshold the image edge extraction, and then for theedge to the appropriate choice of PSF function, reduce the scope of the edgeoscillation, then combined with wavelet filtering, and further to the edge noisereduction; thirdly, for the defects of the POCS algorithm, such as slow convergenceand the weak noise reduction capabilities, in this thesis, on the basic of the POCSalgorithm combined MAP algorithm, and made some simplified MAP algorithm toimprove the ease of programming.In order to verify the effectiveness of the improved method, two sets ofexperiments were made; the experimental results show that the algorithm has a goodeffect.