【作者】 李成龙；

【导师】 尹辉；

【作者基本信息】 北京交通大学 ， 计算机技术， 2012， 硕士

【摘要】 图像增强技术是指通过特定的方法突显图像中某些想要的信息,抑制或消除无关信息,以改善图像的视觉感知效果。经过处理后的图像更加适合人眼观察或计算机分析处理。图像增强的目标和方法往往与具体应用相关,研究人员通常针对不同采集条件下的图像特性以及不同的的应用任务要求设计对应的图像增强算法,很少有通用的增强算法。近年来由于场景和成像设备的限制,在不均匀光照和低光照条件下拍摄的图像视觉效果较差,人眼难以识别,机器难以处理,急需对其进行增强处理。人类的视觉系统具有高度完善的自适应调节功能,能够在多种不同光照条件下准确地感知外部场景。因此,模拟人眼的处理过程来研究图像增强技术是一个有效的途径。本文对现实场景中的不均匀光照和低光照条件下采集的图像特性进行了深入研究,提出了基于照射/反射模型的图像增强算法。主要工作包括：1.研究了照射/反射模型,针对现有算法在照射分量估计上的不足,提出了基于分块的照射分量估计算法,并在真实数据集上验证了基于分块估计照射分量的必要性,为后续的图像处理奠定基础；2.研究和分析了照射/反射模型相关算法在动态范围压缩上存在的不足,提出了基于分段对数函数的动态范围压缩算法。在真实数据集上的实验结果证明算法能够有效地增强亮度较低的区域,抑制过亮的区域,使图像符合人眼的视觉特性；3.针对传统的照射/反射模型中的色偏和饱和度不够等问题,提出了基于灰度世界假设的色彩校正算法。在真实数据集上的实验结果证明算法能够有效地消除图像增强后由于不良光源引起的色偏现象；4.将上述算法应用到扣件状态检测系统中,通过在真实数据集上的实验结果证实,本文算法能够提高扣件的定位准确率和识别能力,同时验证了本文算法在去除不均匀光照和低光照影响的有效性。更多还原

【Abstract】 Image enhancement is a kind of technology about highlighting some information which you want, and removing irrelevant information which you do not need, so that it can improve visual effects and perception. Images are more suitable to the human eyes and computer recognition after the enhancement. The goals and methods of image enhancement are often related to the specific applications. Researchers commonly use a specific algorithm for different images under various conditions. There is few common enhancement methods for enhancement. In recent years, images can not be recognition by the restrictions of the scene and image equipment, even under the uniform illumination and low lights conditions. Human visual system can be able to accurately perceive the external scene in a variety of light conditions with the adaptive adjustment function. Therefore, the way to simulate human eyes is an effective enhancement method. In this paper, the images under uniform illumination and low lights have been researched. An enhancement algorithm based on illuminance/reflectance model is proposed. The main tasks of this paper include:1. Illuminance/reflectance model has been studied. Estimate illuminance by local of image could more accurately, so illuminance estimation based on image block is proposed in this paper. The necessity of this algorithm has been verified in Database.2. Image enhancement algorithm on dynamic range adjustment and contrast enhancement has been studied. Dynamic range adjustment in illuminance/reflectance model Based on logarithmic function was proposed. According to the database, the algorithm can effectively enhance the brightness of the lower region; inhibit the brighter region, so that the image will be more suitable for human eyes observation.3. The algorithm of gray world and white balance has been studied. A method based on white balance was proposed. This algorithm can effectively remove color cast for visual observation of human eyes.4. The proposed enhancement algorithm is used to fastener detection. It has improved accuracy of fastener position and identification of fastener position.更多还原