Improved Harmony Search Approach based DCNN for Image Restoration Model

Abstract

: In various fields, image restoration has received huge interest and many researchers introduce several image restoration techniques to restore hidden clear images from degraded images. Moreover, aforesaid approaches performances are estimated impartially remnants the huge confront that might delay the furthermore improvement of developed image restoration techniques. Hence, an efficient noisy pixel prediction on the basis of the image restoration is introduced that uses the Deep Convolutional Neural Network (DCNN) classifier to restore the input image from several noises, such as random noise as well as impulse noise. An Improved Harmony Search Algorithm (IHSA) is adopted to train the DCNN optimally based on minimum error. After identifying the noisy pixels, by exploiting the neuro-fuzzy system the enhancement of pixel is performed. Finally, the experimental analysis is performed and the image restoration performance on the basis of IHSA is analyzed based on the SDME, PSNR, and SSIM. Ultimately, the adopted model attains the maximum PSNR SSIM for images with random noise, as well as maximum SDME with impulse noise, correspondingly.