Research On Infrared Dim-Small Target Super-Resolution Restoration Algorithm Based On Improved POCS

With the progress of science and technology, infrared imaging technology is widely applied to industry, agriculture, medicine and other fields. Comparing with visible light imaging, infrared imaging has the strong ability to penetrate smoke and dust, and can work day and night. Therefore, it becomes the research hotspot. Infrared imaging detects and identifies the target by receiving the infrared radiation, these targets include aircrafts and missiles, they may also be vehicles and tanks on the ground. However, when detected from a long distance, a target in the image is usually only a few dozens or even a dozen of pixels. Furthermore, it also affected by the atmospheric impurities and the restriction of the infrared detector sensitivity. Eventually, it is difficult to extract the target due to its merging in the background clutter on the infrared image.This thesis restores the dim-small target image in order to facilitate the further detection and identification of the target. Among so many restoration methods, the super-resolution restoration is a very effective method, which can restore the information beyond the system cutoff frequency and provide more detail information. The premise condition is that we need to get a series of low resolution images in the same scene. Each low resolution image represents different aspects of the scene. Super-resolution restoration gets the motion estimation of the relative movement between the low resolution images. And then the information on the high resolution mesh is fused to get a high resolution image. Its core idea is to get spatial resolution using temporal bandwidth. Restoration of the dim-small target image can make the target much more clear and "separate" from the complicated background clutter, which is helpful for further extraction and target recognition.POCS is a classic algorithm among super-resolution restoration algorithms. It characters for fast convergence rate and better edge retention. Therefore, this thesis uses the POCS algorithm for restoring infrared dim-small target image, and puts forward the improved algorithm.As POCS super-resolution reconstruction algorithm iteration has a long running time, this thesis puts forward a new POCS super-resolution reconstruction algorithm based on region selection. The algorithm selects the possible target points firstly according to the histogram of the image, and then divides the image into target region and background region. Then the target area is restored by super-resolution restoration. The proposed method can save the operating time.In addition, the POCS algorithm estimates the initial high resolution image using the bilinear interpolation, which has low complexity. However it will lead to the edge blur, which is not conducive to further motion estimation. In this thesis, the Newton interpolation algorithm and the POCS algorithm are combined. The gray value of the pixel has strong correlation with the source pixel in Newton interpolation algorithm, accordingly, it can effectively improve the edge fuzzy and also obtain a better result.