Super Resolution Image Reconstruction using Geometric Registration

Super-resolution algorithms are facing a number of issues, in that image registration is one of the major challenges. Here, we need to generate an improved resolution image that is reconstructed based on several geometrically warped, linearly blurred and noisy images. The ultimate goal is to compute a high-resolution image from low resolution image. For this an approach which makes use of geometric registration, feature detection and Contrast invariant feature transformation with affine invariant region descriptors is proposed in this paper. Geometric registration is used to align an improper image in to proper image. Feature detection is used to detect edges and corners of an image. The CIFT stretches image contrast by increasing the intensity value of each pixel to obtain high resolution image. The proposed approach provided a super resolution image with appropriate shape, size and contrast, and increased accuracy and reduced processing time. Keywords— Super-resolution, Low resolution, High resolution, Geometric registration, Feature detection, Contrast Invariant feature transformation, Affine Invariant Region Descriptors.