摘要
对于一个溺水报警系统.为了将游泳者从背景中分离出来以便分析其运动特征.需要对受各种噪声污染的观测图像进行恢复处理。由于泳池水下图像受水波干扰较为严重.信噪比低.直接使用传统的图像恢复方法对其进行处理.效果较差。本文根据鲁棒估计方法抗噪性能强的特点.将其与图像恢复技术结合起来.提出了一种自适应鲁棒平滑滤波算法。通过泳池水下图像处理实验证明,该算法可以在充分去除噪声的同时.很好地保持原始图像中大部分的边缘结构。因此.本方法具有一定的实用价值。
To a drowning warning system, the first thing we have to do is to separate swimmer from the background. This needs us firstly to restore the images from all kinds of noises. For the underwater image of swimming pool, its S/N is low and the edge is fuzzy. If use traditional image restoring method to dispose it directly, the result is not expected, existing weaknesses such as enlarged edge, denoise incompletely and so on. In this paper, according to the high stabilization of robust estimation to noise, we combine robust estimation technology and image restoration method together, and propose an adaptive robust image smoothing algorithm Experiments with underwater images of swimming pool prove that the proposed approach can preserve the structure information in the original image while smoothing it simultaneously. Thus, the proposed approach is valuable in practice.
出处
《仪器仪表用户》
2008年第1期53-54,共2页
Instrumentation
关键词
鲁棒估计
水下图像
图像平滑
尺度参数
robust estimation
underwater image
image smoothing
scale parameter
作者简介
蔡小秧(1983-),女,浙江温州人,北京工业大学电子信息与控制工程学院硕士研究生,主要研究方向:智能图像处理和模式识别
陈文楷(1946-),男,北京人,教授,北京工业大学电控学院硕士生研究生导师,清华大学特聘教授.国务院学位委员会委员.主要研究方向:数字图像处理、神经网络和遗传算法等。