基于神经网络的低照度真彩色图像增强  

True color image enhancement based on neural network under low illuminance

作　　者：窦易文[1] 周鸣争[1] 卢桂馥[1] 

机构地区：[1]安徽工程科技学院计算机科学与工程系,安徽芜湖241000

出　　处：《计算机工程与设计》2010年第12期2835-2837,2841,共4页Computer Engineering and Design

基　　金：安徽省高校青年教师科研资助计划基金项目(2006jq1156)

摘　　要：针对人类在低照度环境下的视觉难以分辨目标形态及其颜色的缺陷,提出了一种真彩色图像增强方法。通过普通摄像机分别采集理想光照环境下和低照度环境下的图像,送入神经网络进行误差反传,修改权值并完成非线性映射,测试过程采用主观和客观的评价方法。与直方图均衡化方法相比,基于神经网络的低照度真彩色图像增强方法有效且能得到了较好的增强效果。In view of the defects that the shape and color of the object are always very difficult to distinguish under low luminance,a novel method is presented to deal with those problems.Using CCD camera,the images under natural illumination and low luminance environment are captured respectively,and all the samples would be sent to neural network.There,before the network generalized,would pass three steps-error propagation,modifying the weights and achieving the nonlinear mapping.The simulation testing methods are used subjective judgment and objective judgement respectively.Compared with histogram equalization,experiments indicate that the presented method is an effective method and show promising enhancement results.

关 键 词：低照度 BP神经网络 彩色图像增强 直方图均衡化 峰值信噪比(PSNR) 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]