A method for smoothing the blur images by deblurring them using GAN is discussed in the paper. The method includes learning techniques to detect the type of filters applied to the image, all possible edges present in the image and counting all those details into account. This method makes assumptions about the filters on the image and uses techniques to de-filter all the affected pixels. This method can also handle the unconstrained blurs caused due to overexposure, underexposure, shaking of the camera lens, zooming of pixels and optical noise. Due to these unconstrained measures of the photograph, the image becomes blurred or semi-blur. The proposed method also focuses on constrains of prior information about blurring filters and parametric definitions of filters used. This method detects all possible blurred edges in its own way using the contrast of all blurry areas. The colour contrasts of blurred images are different from that of deblurred images. Also, it can work on both single-frame and multi-frame scenarios. The constrained parameters of the blurry areas in multi-frames are used in achieving improved quality images.