Optimized Weight Maps and Fusion for Underwater Image Enhancement

摘要

The significant growth in technology advancement leads to underwater videos and images capturing for different purposes. However, the captured underwater images or videos suffer from the low contrast, color distortion, and haziness, thus it is required to enhance the quality of such images or videos for further analysis. The image enhancement techniques already proposed, however such methods are not applicable for the underwater images with the different physical properties. It is challenging research problem to optimize the underwater image quality. In this paper, first attempt towards the restoration of underwater image using the fusion based approach proposed. Previously, the weight maps are computed and fused to enhance the quality of images, but the weight maps introduces the artefacts while performing the fusion, hence to overcome that problem, the optimized fusion technique of weight maps for underwater image enhancement designed. The multi-step fusion approach works independently on two derived images from the original image. Then to optimize the visibility of underwater image, the three weight maps such as saliency, luminance, and chromaticity computed. These weight maps are fused in multi-step manner to overcome the challenge of artefacts and generate the final restored underwater image. The results prove the effectiveness of proposed model compared to single step fusion technique.