Computationally Efficient Super Resolution Algorithm
Abstract
Super resolution image can be obtained from combining several low resolution noisy and blurred images. We propose an efficient algorithm to produce super resolution microscopic images. In the proposed algorithm, accurate sub-pixel motion between images is essential for reconstructing the image. Denoising is carried initially by adjusting the low resolution images. Shift fusion approach is applied to enhance the resolution of image and improved optical flow method is used for registration of images. The proposed method is applied to each color channel separately. The results are tested with synthetic downgraded images, popular low resolution datasets and experimental real-life images showing significant improvement in quality of images, with considerable less time cost and memory consumption than those of existing methods. Qualitative analysis is studied through edge detection method and observing visible features. Quantitative analysis is inspected showing improvement in resolution by measuring observable minimum distance.