【作者】 韩涛；

【导师】 闫成新；

【作者基本信息】 中国石油大学 ， 机械工程， 2010， 硕士

【摘要】 我国是个海洋大国,拥有1.8万公里长的海岸线和300万平方公里的海洋国土,对海洋的开发和海洋规律的探索对我国的国民经济和军事都有重要的意义和价值。图像信息是人类获取信息的主要来源,但由于水下光学成像系统受到水下成像环境的限制,低质量的水下图像直接影响了人类对海洋的探索。本文从数字图像处理的角度,对水下图像预处理技术进行研究,以提高水下图像的质量。我们采取了大部分的图像预处理方法用于实际水下图像的实验分析,并且针对不同的处理方法,进行了改进尝试。我们在局部测度增强方法中提出了基于局部复杂度的水下图像增强方法,该方法对于处理尺寸较小的水下图像具有较好的处理效果;针对水下图像的亮度分布特点,对小波变换增强方法进行了改进,提出了基于小波和非线性迭代的水下图像增强方法,该方法既能增强水下图像又能较好地处理水下图像的亮度缺陷。本文分增强和降噪两部分对各方法进行了比较分析,得出了水下图像处理中应该优先选用的方法。在水下图像增强中,如果是亮度较均匀但整体亮度偏暗的水下图像,可选择同态滤波(低频增益设为大于1)、基于小波和非线性迭代的增强方法;如果是亮度不均匀的水下图像,优先选择基于小波和非线性迭代的增强方法。在水下图像降噪中,如果是椒盐噪声,则选用中值滤波;如果是高斯噪声,则选用贝叶斯小波阈值收缩法。更多还原

【Abstract】 Our country has 18,000 kilometers long coastline and 3,000,000 square kilometers sea national territory, so the studies on seas’rules, as well as seas-exploration are vital for our country’s economy and military. The images’information is the basic and significant source for human to obtain information. However, the poor underwater condition results the underwater images with a too low-quality for the seas-exploration. From the perspective of digital images’processing, we did some research on the underwater images’pre-processing methods to improve their quality.We adopted the majority of existing images’pre-processing methods, and had ameliorated some of them to process the real underwater images. In the local measure enhancement method, we proposed a new enhancement method based on the local complexity, which was effective for the small size underwater images. Aiming at the brightness distributive characters of underwater images, we ameliorated the wavelet transformative enhancement, and proposed a new enhancement method which was based on the wavelet and nonlinear iteration. This new method will not only enhance the underwater images’effectively but also solve the problem of the underwater image’s uneven brightness distribution.We could choose the prior method in the certain condition after comparative analyzing all those methods from both enhancement and de-noised parts. For the underwater images’enhancement, if the image’s brightness is even, but for the whole it is dark, we should choose the homomorphic-filtering method(low-frequency gain is supposed to be bigger than 1) and enhancement method based on wavelet and nonlinear iteration. If the underwater image’s brightness is uneven, we suggest choosing the enhancement method based on wavelet and nonlinear iteration. In the process of underwater images’de-noising, if it is pepper-and-salt noise, we’d better choose median filtering; if it is gaussian noise, we’d better choose BayesShrink de-noising method.更多还原