Most of the image acquisition algorithms neglect the illumination problem like shadows and direction of illumination changes as well the image degradation caused by motion blur, noise introduction in several intermediate process. This paper presents a novel method of high resolution image generation by interpolating local gradient field and subsequent training of LR cluster patches by fuzzy learning. The query image, after several iterative resolution steps with help of fuzzy clustering, a penalty approach is associated as if a feedback path to eliminate the overestimation of HR pixels. A sparse monitoring for removing aliasing of samples is formulated to remove the gradual noising of pixels.