海面目标场景红外图像增强算法研究

发布时间：2018-01-05 21:42

本文关键词：海面目标场景红外图像增强算法研究　出处：《大连海事大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来,随着我国海上运输业和旅游业的发展,海上遇险事故频频发生,为了最大限度的减少事故损失,保障人民生命财产安全,及时有效的海上搜救工作至关重要。海上遇险事故常发生在夜晚或者大雾、大风浪等恶劣气候条件下,如果采用比较原始的搜寻手段进行海面遇险目标搜寻,不仅费时费力,而且效率极低。鉴于红外成像技术具有透过烟雾能力强、可昼夜工作、抗干扰能力强等优点,红外成像系统已成为最适用于全天候海面目标搜寻的实用系统。但是,由于受到红外探测器材料、成像机理及环境等因素的影响,红外图像往往具有对比度低、细节模糊等缺点,这会对图像的观察、特征提取和目标检测等工作产生严重影响。本文主要研究海面目标场景红外图像增强算法,结合海面目标场景红外图像的特点,联合使用两种红外图像增强算法,提高红外图像对比度,增强细节,减小模糊。本课题来源于国家科技支撑计划课题"深远海海面遇险目标机载红外搜寻定位技术"。通过红外搜寻系统获得的海面目标场景红外图像的对比度很低,图像整体偏暗,几乎无法从原始图像上得到目标的任何明显轮廓信息。本文在研究了常用图像对比度提升算法的基础上,利用引导滤波将图像分为基础子图像和细节子图像,然后采用基于最大熵的双直方图均衡方法处理基础子图像,采用Gamma变换增强细节子图像,提高图像整体对比度,增强细节,改善"过增强"现象。由于受到海面雾天雨天等恶劣天气、远距离传输和大气衰减等因素的影响,对比度拉伸后的图像依然存在细节模糊问题,本课题采用基于大气散射模型的方法实现红外图像的模糊去除。该方法利用四叉树法估计大气光值,通过定义代价函数,自适应估计透射率;采用引导滤波优化透射图来消除图像分块后的人工痕迹;通过大气散射模型反演得到增强后的红外图像。为了验证所采用的算法的有效性,本文进行了大量实验,并对算法处理后图像进行客观质量评价,从图像的视觉效果和图像客观质量评价两方面验证了算法的有效性。
[Abstract]:In recent years, with the rapid development of China's shipping industry and tourism, maritime distress accidents occur frequently, in order to minimize the loss of accidents, protect people's lives and property safety is very important, timely and effective maritime search and rescue work. Maritime distress accidents often occur at night or adverse weather conditions of fog, storm and so on, if by comparison the original search means distress target surface search, not only time-consuming and low efficiency. In view of infrared imaging technology is through the smoke ability, can work day and night, strong anti-jamming capability, infrared imaging system has become the most practical application system to target all-weather surface search. However, due to material by infrared detector effect, imaging mechanism and environmental factors, the infrared image has low contrast, fuzzy details and other shortcomings, it will observe the image feature Serious impact on Extraction and target detection. This paper mainly studies the target scene infrared image enhancement algorithm, combined with the characteristics of infrared image of the target scene, the combined use of the two infrared image enhancement algorithm, improve the contrast of infrared image, enhance the details, reduce the blur. This topic is from supporting project "far-reaching sea distress target airborne infrared searching and positioning technology in the national science and technology. Through infrared search system contrast surface target IR image scene is very low, the whole image is darker, almost impossible to get any obvious contour information of the target from the original image. Based on the study of the commonly used image contrast enhancing algorithm based on the filter will guide the image is divided into basic sub image and sub image, and then using the maximum entropy method of histogram equalization processing based on sub image based on the Gamm A transform sub image enhancement, improve the overall image contrast enhancement, details, improve "enhanced" phenomenon. Due to the sea fog rain and other inclement weather effects, long distance transmission and attenuation factors, image contrast stretching after details still exist ambiguity, this paper adopted the method of atmospheric scattering model realization the infrared image based on fuzzy removal. This method uses four fork tree to estimate the atmospheric light value, through the definition of the cost function, adaptive estimation is adopted to guide the optimization of transmission; transmission filter to eliminate image artifacts after blocking the infrared image enhancement; after the inversion of atmospheric scattering model. In order to verify the validity of the algorithm in this paper, a lot of experiments are carried out, and the algorithm of image quality assessment, evaluation of two aspects of verification from the visual effect of image quality objective The effectiveness of the algorithm.

【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】