用于数字监控系统的图像去噪算法的研究与实现  

Research and implementation of denoising algorithm for digital monitoring system

作　　者：彭波[1] 崔永普[1] 吕小晴[1] 

机构地区：[1]中国农业大学信息与电气工程学院,北京100083

出　　处：《中国农业大学学报》2004年第5期62-66,共5页Journal of China Agricultural University

摘　　要：针对数字监控系统因监控环境亮度不足导致监控图像存在大量噪声的情况 ,分析了图像的主要噪声源及其特点和性质 ,确定影响图像质量的主导噪声为CCD摄像管产生的读出噪声。比较了各种常见的图像去噪算法 ,在综合考虑图像去噪平滑效果、图像清晰程度和时间复杂度的基础上 ,提出将数学形态学法与中值滤波法相结合的用于数字监控系统的图像去噪增强算法。仿真试验及对多幅在不同亮度环境下拍摄图像去噪前后效果的对比结果表明 ,使用数字形态学滤波与 3× 3中值滤波结合的算法去噪 ,可将压缩后的录像码流数据量降低 15 %～2 0 % ,且去噪后的图像比较清晰平滑。由于算法的时间复杂度较低 (为O(N) ) ,速度较快 ,可满足实时性处理的要求 ,所以可将其直接移植到数字监控录像系统中。试验结果验证了本算法的可行性、实时性和鲁棒性。The main noise sources and characteristics of monitor images for many noises of digital monitoring system were analyzed in the monitoring situation of insufficient illumination. The main noise affecting the qualities of monitor images was read-out noise produced by the charge coupled devices. The common denoising algorithms were compared, and the enhanced denoising method was designed combining mathematics morphology with median filter after balancing image smoothness, clearness and time complexity for digital monitoring system. Through making a lot of emulational experiments and comparing many images of before and after denoising in different illumination environments, it was expressed that the compressed video stream decreased by 15%-20%, and the images become clearer and smoother using the enhanced denoising method combining with mathematics morphology with 3×3 median filter. This algorithm had lower time complexity (O(N)). So it could be used in digital monitoring system without influencing the real-time property. It was also confirmed that the algorithm was feasible, real-time and robust.

关 键 词：噪声 数学形态学法 中值滤波 去噪 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构] TP311[自动化与计算机技术—计算机科学与技术]