Research On Super-resolution Reconstruction Algorithm For Sequence Images

As people increase requirements of image resolution, the resolution of imaging devices often do not meet the requirements. As the inherent sampling frequency of imaging senor limits image resolution in the process of obtaining digital image. Obtaining higher resolution images from the hardware requires high cost, so how to use of economically feasible software method to improve image resolution received extensive attention. Using information of sequence low-resolution images for super-resolution reconstruction can get more satisfactory high resolution images. Currently super-resolution reconstruction technique has become the focus of the field.This paper mainly studies the estimation of motion parameters between sequence images and super-resolution reconstruction algorithm. For estimation of motion parameters,this paper using block motion estimation algorithm, studied and implemented a variety of search algorithm to find the best matching block to estimate the motion parameters. However, this algorithm can not estimate the motion parameters to sub-pixel and accuracy is not high enough. So this paper studied and implemented a variety of sub-pixel registration algorithm and proposed two improved algorithm, they are improved sub-pixel image registration of based on cross-correlation and based on Lucchese and aliasing. The experimental results show that, sub-pixel image registration algorithm of based on cross-correlation ensuring the accuracy,while greatly reducing the time required and improved sub-pixel image registration algorithm of based on Lucchese and aliasing estimated motion parameters have more accury.For super-resolution image reconstruction, the paper studied POCS method in depth. Analysised the reason of edge defect of high-resolution image reconstructed by basic POCS method and proposed POCS super-resolution image reconstruction algorithm of improving the quality of the edge. Introduced the butterworth low-pass filter in iteration process to smooth reconstruction image and weaken the edge of boundary, so edge can better adapt to convex and improved the quality of edge of reconstructed image. The experimental results show that, the reconstructed image quality of improved algorithm is better than the reconstructed image quality of the basic POCS algorithm.