Research On Infrared Subpixel Image Restoration Technology

With the development of aerospace technology, people have higher and higher requirements to remote sensing image resolution, the direct solution to improve the image resolution is to improve the size of CCD detector, but this method is not only difficult in technology but also costs a lot, in this case, subpixel imaging technology becomes a new choice. subpixel imaging uses mutual misalignment CCD detector to photos the same object, then through using software algorithm to recover the low resolution image.Firstly in this paper introduces the principle of infrared subpixel imaging technology, CCD detector arrangement, and analysis the MTF of imaging system, then has a key discussion about a high precision MTF edge measurement technology, in lab through moving the knife-edge to directly observing the change of a single pixel grayscale value, uses three methods to fit ESF, namely Linear function, Gaussian function and Fermi function,and compare the fitting effects, at last, calculates the MTF by Fermi fitting, which has the best fitting effect.comparing with the traditional method which obtains ESF from remote sensing image, the method is simpler and more intuitive.At the same time, we have a pretreatment to the sub-pixel image, according to the image of the dead pixel phenomenon, using different interpolation methods to remove single line and double line dead pixels. Infrared nonuniformity correction is also an important aspect, uses global non uniform algorithm and weighted mean filter algorithm to correct, the former method has the better correction effect. The difficulty of subpixel image restoration is registration, using manmachine interactive registration method to register and fuse image, the advantage of this method is that the registration parameters can be calculated by several data points which are selected from the image.In the process of image restoration, based on MTF and reconstruction,using the two methods to restore image.when measure system’s MTF by new method, respectively by inverse filtering, Wiener filtering and constrained least squares filtering algorithm for a complete image restoration, then evaluate the restored image quality by objective evaluation,the test results show that, restoring image by constrained least squares filtering algorithm has the best effect; In the restoration based on reconstruction, respectively by Interpolation method and POCS algorithm, Interpolation methodcontains nearest interpolation, bilinear interpolation and cubic interpolation, the traditional POCS algorithm has some defects such as image shock, so this pape develops a new method to improve,increases a new process to remove the image noise in the initial low resolution image, the results show that the quality of the reconstructed image restoration is obviously improved, the method based on MTF image restoration has a strong capability to remove noise, but has less restored image details, the method based on POCS algorithm can obtain more image details, but has a poor capability to remove noise.