Research On Super-resolution Reconstruction Of Video Sequences

With the rapid development of multimedia technology in recent years, the demand for high-quality video images grows fast with people. A high-quality image always contains further detailed information of targets, and it is of great value for analysis and post-process. However, because of physical limitations and cost issues of sensor, it makes difficult to obtain high-resolution images through hardware improving. So it a natural choice when people turn to the software method of image super-resolution reconstruction technology.Super-resolution reconstruction technique refers to restoring a high-quality and high-resolution image from multiple low-resolution video sequences which are degraded by warping, blurring, noise and aliasing. There are two different researching methods:frequency domain method and spatial domain method. It can enhance the spatial and temporal resolution of images effectively without any upgrade of current equipments. Thereby, it is worthy of notice both for academic studies and applications.At first, this paper reviews the researches and applications of the video image super-resolution reconstruction in now society, and compares the major algorithms. Then, several improved algorithms of video image super-resolution reconstruction are given with good experimental results. Topics included are as follows:Ⅰ. Video image super-resolution reconstruction theory. It introduces the model of degraded imaging, the basic theory of super-resolution reconstruction, the quality evaluation of a reconstructed image, and then compares the performances of the frequency domain algorithm and space domain algorithm.Ⅱ. Based on motion estimation algorithm of video image super-resolution reconstruction. Obviously, the key to the super-resolution reconstruction is a good motion estimation result. Motion estimation algorithm based on block is proposed with the difference idea introduced to block searching algorithm, it can improve the search accuracy performance; Motion estimation algorithm based on spatial and temporal gradients is developed for other more general videos. It presents a motion estimation algorithm based on the Horn-Schunck optical flow estimation method. Through selecting the gradients adaptive to reduce the occlusion problem in computing optical flow. The experiment results show that the two improved algorithms make better image quality.III. The MAP-based adaptive video image super-resolution reconstruction algorithm. Multi-filter function is applied and the new regularization function is used to extract images texture. The new regularization function could better preserve slope and roof edge while keep global smoothing. Finally, the experiment results show that the reconstructed image quality is better than other MAP-based algorithms.IV. The summary. There are some ideas and prospects of video sequences super-resolution reconstruction.