Under Water Image Intensification Based on CNN Using Deep Learning

Image enhancement inside the water and reconstruction is becoming a challenging task and has gained priority in recent years, as the human eye cannot clearly perceive underwater images.Due tothe absorption andscattering effect on light when travelling in water, underwater images exhibit color deviation, low contrast and blurry details. The existing methods, such as supplementary information- based methods, nonphysical model-based methods and physical method-based methods rely on polarization filters, modify image pixel values and assume the attenuation coefficients to improve visual quality. These methods may give unstable results and provide less accuracy. The main objective of the proposed system is to improve the quality of the image by adjusting the contrast, illumination and enhancing the pixel edge are applied to an underwater image.White Balance algorithmis used to correct the color casts, Histogram Equalization is used to improve the contrast and Gamma Correction controls the overall brightness of an image. We have used the Convolution Neural Network algorithm for image classification and trained all the images from dateset based on three techniques to improve overall accuracy. Index Terms -Image Preprocessing. White Balance, Histogram Equalization. Gamma Correction and CNN.