Research on the influence of initial value estimation on image super resolution reconstruction

Abstract

In order to improve the quality of images, an image super resolution reconstruction algorithm based on initial value estimation is proposed. In iterative back projection (IBP) algorithm, the reference image is restored through Wiener filtering. Bilinear interpolation is performed on the restored image, other than interpolating the reference image directly adopted in the traditional methods. The generated image is taken as the initial value of IBP algorithm. Image super reconstruction is performed on multiple simulated low resolution images. The experimental results shows that the initial value plays an important part in image super resolution reconstruction algorithm. The reconstructed image achieves higher peak signal to noise ratio (PSNR) and better visual effect than that in traditional way. The IBP algorithm in this paper has fast convergent speed and small amount of calculation. This initial value estimation method may be extensively used to other iterative solving methods, such as project onto convex set (POCS).