Underwater Image Enhancement Algorithm Based On Image Fusion

Underwater images is one of the important ways to obtain the information of the underwater world, but the underwater images degrade seriously due to the blur of detail,contrast decrease and color distortion, etc. Underwater image degradation is the primary result of light attenuation and scattered by underwater medium. The light is absorbed by water, which leads to the color attenuation of images. Scattering is divided into forward scattering, which results in the blur of detail, and backward scattering which causes the decrease of contrast. So the main task of the underwater image enhancement techniques is to eliminate the image degradation, enhance image clarity and restore the true color in the process of underwater imaging. Based on the reason of underwater image degradation, this paper adopt an underwater image enhancement approach based on fusion techniques. The main work of this paper includes:(1) According to the feature of color distortion, white balanced algorithm is used to correct the color of the original image. And for the reason of contrast degradation, contrast enhanced algorithm is used to obtain brightness enhancement of the original image. Weight map of region of interest (ROI) of two image is adopted to increase the visibility and to render the details.(2) Two region of interest weights of color corrected and contrast enhanced versions were defined to be fused. And three weight maps of brightness, chroma and saliency of each version are obtained, after the three weight maps normalized, three new weight maps of each version can be obtained. Then, the three weight maps are summed to one weight map for the final weight map. Finally, the two versions are weighed by using the final weight map to obtain ROI color corrected image and ROI brightness enhanced image.(3) The method of multi-resolution fusion is adopted to process two ROI images. First, wavelet transformation for decompose two ROI images are used to generate low-frequency and high-frequency components. Then, the weighted average for fusion low-frequency components and the local variance method are used directly for fusion high-frequency components. Finally, wavelet inverse transform for reconstruction image are used to realize image enhancement.The experimental results demonstrates that the algorithm in this paper improves the image clarity. Our approach arrives a better result compared with other approaches of enhancing underwater image, and also shows the promotion effects in feature point detection in underwater image.