Pretreatment And Analysis Of Images Of Underwater Historical Sites

In recent years, with the increasing development of ocean exploitation, exploring the ocean is no longer limited to military purposes, and is more inclined to civilian purposes. Activities for observation, understanding and exploration of ocean involve a great deal of processing and analysis of visual information. Hence, preprocessing and analysis technology of underwater images is seized of important status in future ocean engineering.Because of the complex underwater environment, the underwater images have their own property. For example, the low contrast, the fuzzy feature, and so on. So it is necessary to preprocess the underwater images before higher steps (vision processing). The author firstly uses the fuzzy enhancement. According to the characteristic of the underwater image, a statistics-based method is proposed to choose the adaptive cross point. Comparing with the method based on maximal fuzzy entropy information, the new one is much faster and works well. It can be also used in real time system. Secondly, the fields of relic's contour are extracted directly by threshold. Thinning algorithm and the pruning method for eliminating short branch in the next step are used for complete contour.The sand hides most of the relic's contour. So the extracted contour is usually just a part of the whole contour. And they are often distorted. First, a low passed filter is used to smooth the contour. And a polygon approximating method based on area error is used to approximate the contour. The turning function is involved as the descriptor. By searching the nearest distance between some parts of the model's turning function and the contour's turning function, it is able to judge whether the contour is a part of the model. The method makes use of the global information of the contour, and can be adjusted by entering parameters.