A New Approach for Super resolution by Using Web Images and FFT Based Image Registration

TL;DR

This paper introduces a novel super resolution method combining FFT-based image registration and sparse reconstruction, improving image accuracy by effectively aligning and enhancing low-resolution images using local descriptors and similarity metrics.

Contribution

It presents a new FFT-based image registration algorithm and integrates it with sparse super resolution techniques, enhancing accuracy over traditional methods.

Findings

Improved image resolution as shown by PSSNR and SSIM metrics.

Effective handling of images with different focal lengths and illumination.

Enhanced super resolution quality compared to traditional approaches.

Abstract

Preserving accuracy is a challenging issue in super resolution images. In this paper, we propose a new FFT based image registration algorithm and a sparse based super resolution algorithm to improve the accuracy of super resolution image. Given a low resolution image, our approach initially extracts the local descriptors from the input and then the local descriptors from the whole correlated images using the SIFT algorithm. Once this is completed, it will compare the local descriptors on the basis of a threshold value. The retrieved images could be having different focal length, illumination, inclination and size. To overcome the above differences of the retrieved images, we propose a new FFT based image registration algorithm. After the registration stage, we apply a sparse based super resolution on the images for recreating images with better resolution compared to the input. Based on the PSSNR calculation and SSIM comparison, we can see that the new methodology creates a better image than the traditional methods.