基于光学成像模型的水下图像超分辨率重构

Underwater Image Super-resolution Reconstruction Based on Optical Imaging Model

目前大多数超分辨率重构算法都是针对空域图像进行恢复。考虑到水下光学条件复杂,直接把现有超分辨率重构算法移植到水下,无法对光线在水中传播产生的散射和衰减进行有针对性的处理。基于此,本文结合不同成像模型,提出一种基于光学成像模型的水下图像超分辨率重构算法。首先,考虑到水中光线散射对图像降质严重,根据水下光学成像模型,利用暗通道原理估算观测数据中的散射光和光线透射图,作为噪声估计的结果;然后对去除散射光的观测数据进行凸集投影超分辨率重构,得到高分辨率图像;最后利用透射图对高分辨率图像进行光线衰减补偿,去除由于水对光线的吸收作用造成的图像亮度降低和模糊,得到最终的恢复图像。对本文算法进行仿真,并与经典的超分辨率重构算法比较,验证了本文算法对恢复图像质量的提高作用。

Now most super-resolution reconstruction algorithms are applied to atmospheric picture restoration. Taking the complexity of underwater optical environment into consideration,it is hard to perform direct processes towards scattering and attenuation caused by water simply by transplanting those super-resolution reconstruction algorithms to underwater images. In such condition,an underwater image super-resolution reconstruction algorithm based on optical imaging model is proposed by integrating different imaging models. Firstly,aiming at the severe degradation in image quality caused by light scattering in water,dark channel prior is used,based on underwater optical imaging model,to estimate scattered light and transmission in observed data to produce the result of noise estimation. Secondly,super-resolution reconstruction of projection onto convex sets is performed on the low-resolution image sequence from which scattered light is removed in order to produce a high-resolution image. At last,aiming at overcoming the decrease in intensity and blur caused by water so as to produce the restored image,light compensation is conducted on the high-resolution image using transmission. By comparing the reconstructed images produced by the proposed algorithm with those produced by classical super-resolution algorithms,quality improvement in restored images by our algorithm is proved in algorithm simulation.