Research On Algorithm Of Image Super-resolution Reconstruction

Image super-resolution reconstruction is a new image processing branch, which reconstructs higher resolution images from a sequence of degraded images with lower resolution. Since image super-resolution reconstruction could overcome resolution restriction of a imaging system and improve resolution without changing the hardware of the imaging system, image super-resolution technique has been widely applied to many fields such as high sensitivity digital television, old video squash, video monitoring, remote sensing monitoring, medical diagnostics; bioinformatics feature extraction and pattern recognition. But super-resolution reconstruction is an ill-posed inverse problem. It doesn't have a unique solution existing and has many problems unsolved. The keys of reconstruction problem are accurately estimation of sub-pixel motion of the low resolution images and interpolation, de-noising and de-blurring for the registered image. Traditional interpolation methods are to process an image with regular data. In fact, image registration for low resolution images is irregular. In this paper, image registration and image interpolation of super-resolution image reconstruction algorithm are researched.In this paper, several research contents are as follows:(1) In order to reduce the complexity of the calculations of image registration, a method for estimating the angles with small-sample data is proposed, which can reduce the time of calculation and does not affect the registration precision. This is very meaningful for real-time realizing of super-resolution image.(2) In order to solve the problems that image interpolation based on adaptive steering kernel regression may result in whole gray deviation and some empty pixels in interpolated image, an improved adaptive kernel regression function is proposed and used to image interpolation. It is constructed by introducing the function of the geometry distance of pixels in neighborhood into adaptive steering kernel, which is produced according to the principle of autocorrelation of an image. Experimental results show that the proposed method has a much better effect than interpolation methods based on classical kernel and adaptive steering kernel. The improved kernel regression function for image processing lays the good foundation for super-resolution image reconstruction.(3) The proposed method of image interpolation based on the improved adaptive kernel regression function is used to image super-resolution reconstruction. Experimental results show that the proposed method is effective.