A Novel Regression Architecture for Underwater Image Enhancement

Abstract

underwater image enhancement is an important section in order to construct a more visually appealing image. They do not depend on the image formation. These varieties of methodologies are usually very simple and quicker than other deconvolution approaches.  A novel architecture is used for training and we calculate the parameters including such as PSNR, BRISQUE and MAE values for comparing the results.  However, the underwater image acquired by the camera has low prominence due to haze induced by light reflected from the surface and isolated by the water particles, attenuation of different wavelengths leads to color deviation. In our proposed implementation to enhance the underwater images more by using the Deep learning convolution neural network (DLCNN) with contrast adaptive histogram equalization (CLAHE) by considering the network depth as 10. The DLCNN-CLAHE technique gives better results when compared with VDSR techniques.