融合暗原色先验和稀疏表示的水下图像复原

Combination of Dark-channel Prior with Sparse Representation for Underwater Image Restoration

由于水下图像成像过程中受光的散射、噪声干扰等因素影响,致使图像质量严重退化。为了去除模糊和抑制噪声,改善水下图像质量,该文提出一种融合暗原色先验和稀疏表示的水下图像复原新方法。该方法首先利用暗原色先验理论计算水下图像的暗原色,然后基于稀疏表示理论对暗原色进行去噪和优化,基于改进后的暗原色计算水体透射率和光照强度以计算最终复原结果,可以同时达到去模糊和去噪的良好效果。实验结果表明,提出的方法有效提高了图像的平均梯度和信息熵等图像像素,从而改善了图像的质量。

Due to the influences of scattering of the light and interference of the noise, underwater image quality is always degraded severely. In order to remove the blur and suppress the noise, and improve the quality of underwater image, a novel underwater image restoration method based on the combination of dark-channel prior with sparse representation is proposed. This method adopts the dark-channel prior theory to calculate the dark-channel image at first, and then uses sparse representation to denoise and optimize the dark-channel image. Based on the improved dark-channel image, the more precise water transmissivity and light intensity can be achieved to compute the final restoration result, effectively eliminating the image blur as well as noise. The experimental results show that the proposed method can effectively improve the image factors, such as average gradient and entropy, so as to compensate the degraded image.