Super Resolution Based Underwater Image Enhancement by Illumination Adjustment and Color Correction with Fusion Technique

Abstract

In underwater photographs are look like low-quality images, the main reason is behind that due to attenuation of the propagated light, absorption and scattering effect. The absorption significantly reduces the light energy, while the dispersion causes changes in the light propagation path. They result in foggy appearance and degradation of contrast, causing misty distant objects. So, for getting the most effective result from that type of image, there must be an enhancement technique that has to be applied. We propose an efficient technique to enhance the images captured underwater by applying a fusion-based technique using super-resolution. For enhancing images, we have followed two steps. The first one illumination adjustment and another one is color correction. Then fusion technique is applied to the resultant image from illumination adjustment and color correction as two inputs and combined them with their maximum coefficient value and received output from there. After that, on the fused output image, we used the Super-Resolution method. In the Super-resolution procedure, low resolution and high-resolution images are used then a bicubic interpolation algorithm and finally, VDSR (very-deep super-resolution) neural network has been used to get the most effective result from an obscure underwater image. For getting the most effective result from an obscure image, a new high-quality and efficient image enhancement method has been proposed in this paper.

An illumination adaptive underwater image enhancement method

“…Instead, they enhance visual perception by adjusting image contrast and emphasizing image features. In recent years, a range of effective underwater image enhancement methods have been proposed, including white balance algorithms, Retinex algorithms [1][2][3] , and color correction algorithms [4][5][6] . In recent years, many researchers have proposed improved underwater image enhancement methods based on this algorithm.…”