Novel Hybrid Deep Convolutional and Cycle Generative Adversarial Networks for efficient Image Restoration

Image Restoration is the vital process in image based analysis, as it enhances the quality of the image and assists in achieving a better level of accuracy in the classification process. The image restoration is performed by mitigating the artifacts in the input image, so that the restored images will be rich in quality with an enhanced level of perception, and provides a pushover for classification process. The existing methods of image restoration techniques experiences a major setback of lack of robustness in handling the complex degradation of the input image and falls back in dealing in denoising and deblurring process. To overcome these challenges, this manuscript introduces a novel integrated version of Deep Convolution Generative Adversary Network (DCGAN) and Cycle GAN (CGAN) to perform efficient image restoration artifacts removal process. The proposed work is tested for its denoising and deblurring capability along with the capacity of handling the complex degradations in the input image. The proposed work is tested with various types of input images added with noise and blurring effects and the performance is compared with the state of the art image restoration techniques.