基于局部纹理特征检测的水下光学图像复原算法

Underwater optical image restoration algorithm based on local texture feature detection

由于不均匀光照及水体的吸收与散射,不同水域的图像存在颜色衰减和信息缺失等问题,为了增强其视觉质量,提出基于局部纹理特征检测的水下光学图像复原算法.首先,为反映图像的雾化程度和纹理特征,定义一种改进局部二值编码;然后,通过提出的模板递归思想结合二值编码图多次迭代分别求取R,G和B三通道的全局背景光;其次,根据图像的最小信息损失原理和光学特性,采用相对总变差算子提取图像的有效结构信息估计透射率图;最后,将估算的三通道全局背景光和透射率值代入光学成像模型反演得到清晰化的图像.实验结果表明:该算法能有效提升对比度,校正色差和图像细节等信息,水下彩色质量评价、水下图像质量评价和信息熵值均优于各对比算法,为海洋信息的探索与开发提供了可行的参考.

Due to uneven illumination and absorption and scattering of water bodies,images in different water areas have problems such as color attenuation and lack of information.In order to enhance the visual quality,an underwater optical image restoration algorithm based on local texture feature detection was proposed.An improved local binary pattern was defined to reflect the atomization degree and texture features.Combined with binary pattern graph,the global background light of R,G and B channels was obtained by multiple iterations through the template recursion.According to the minimum information loss and optical characteristics of the image,the effective structure information was extracted by relative total variation to estimate the transmittance map.The estimated three channels global background light and transmittance were substituted into the imaging model to obtain a clear image.The experimental results show that the proposed algorithm can effectively improve the contrast,correct chromatic aberration and image details.The underwater color image quality evaluation,underwater image quality measurement and entropy values are better than the comparison algorithms,which provides a feasible reference for the exploration and development of marine resources.