Study Of Supre-Resolution Restoration For Video Surveillance Sequences

Super-resolution restoration refers to restoring a high-quality and high-resolution image from multiple low-resolution degraded images by the comprehensive utilization of wealthy complementary information between the multiple frames. Under the conditions of the existing hardware to make up for the lack of spatial resolution of the original image, break through restraint of resolution of image acquisition to improve the spatial resolution and image acuity. Super-resolution restoration can fully develop the potential of the existing image data,which illustrate a good prospect of application and extension in pattern recognition,video processing,remote sensing and medical imaging.In the application of video surveillance, imaging equipment can attain repeat images of the same region of the video image, but due to limitation of imaging system and a variety of external environmental factors, these images at a given resolution level, affecting the image visual quality and postprocessing. Compared with other application areas,super-resolution reconstruction of surveillance video is extremely critical because of unique imaging system and extensive prospect. The article focuses on the researche of surveillance video sequence super-resolution reconstruction.Analyze several key elements of super-resolution reconstruction on the basis of the traditional model, in line with these factors to improve super-resolution reconstruction on the algorithm framework of maximum a posteriori super-resolution reconstruction. According to the sampling features of surveillance video images in the data preprocessing stage of super-resolution reconstruction, a high-qualified frame extract method is proposed. Based on an overall consideration of variance of low-resolution images and mean square error between reference image and the low-resolution images, extracting high-quality frames which contribute to super-resolution restoration and excluding those exsisting serious degradation and loss of the critical information frames, against the initial external effects of reconstruction, increased stability of the algorithm and improve the quality of reconstruction.In order to eliminate the influence of image degradation,we improved blind super-resolution reconstruction algorithm based adaptive edge. Image degradation process can be described as motion compensation matrix and the fuzzy coefficient matrix. Motion compensation matrix reflects the geometric deformation, fuzzy coefficient matrix reflects the fuzzy, noise and other factors such as the impact of sub-sampling. To use the blind super-resolution reconstruction algorithm can eliminate images degraded factor to achieve high-resolution reconstruction.According to the movement model features of surveillance video sequence, a super-resolution reconstruction based ground separation is proposed.Algorithm introduced a geometry movement mapping matrix in the observation model, separating the background area and moving object area of video sequence, eliminating the influence of background pixels which close to moving object area in the motion parameter estimation estimation.Thus attain more accurate motion parameters and save time of reconstruction algorithm to keep the edge of the image reconstruction results...