Fusion based Underwater Image Enhancement and Detail Preserving

Abstract

Underwater images suffer degradation of image quality and visibility due to color retention and contrast under non-uniform illumination of light penetrating in the underwater medium. In this paper, we propose a frequency domain model-free method comprising four phases to improve underwater image quality. The first phase includes homomorphic filtering with a high emphasis filter to obtain the edge region by enhancing the contrast of the RGB image. The second phase comprises a discrete wavelet transform (DWT) decomposition of the homomorphic filtered output image from the first phase, and wavelet denoising to obtain the background region by enhancing low-frequency components of each RGB channel. DWT improves the further edge region and enhances the background region. To overcome color distortion and low contrast problems, the underwater image is reconstructed using Inverse DWT and color correction. The third phase is wavelet-based image fusion with the mean approximation of low and high-frequency components. The fourth phase includes a guide filter applied to the fusion output image to preserve the image edges. Further, the unsharp masking filter amplifies the high-frequency components to enhance the edges. The proposed method experimental results effectively improve in terms of low contrast, color retention and better visibility.