This work proposes a method for underwater image enhancement using the principle of histogram equalization. Since underwater images have a global strong dominant colour, their colourfulness and contrast are often degraded. Before applying the histogram equalisation technique on the image, the image is converted from coloured image to a gray scale image for further operations. Histogram equalization is a technique for adjusting image intensities to enhance contrast. The colours of the image are retained using a convolutional neural network model which is trained by the datasets of underwater images to give better results.