基于NSCT变换的随机游走红外图像分割方法
2018-05-15庞春江张锦文
庞春江 张锦文


摘 要:红外图像具有灰度变化不平均、分辨率较低、噪声较多等特征,传统的图像分割方法难以对红外图像分割出有效实体。为解决这一问题,设计了一种基于非下采样Contourlet 变换(NSCT)的随机游走红外图像分割算法。传统的随机游走图像分割方法,图像部分灰度值的不均匀变化可能改变随机游走算法的路径,而且分割目标的轮廓很容易受到图像背景噪声影响,所以可以通过增强图像的目标轮廓,抑制图像噪声,以达到更好的分割效果。NSCT变换是一种非常有效的图像增强方法,首先采用NSCT变换对红外图像进行多方向、多尺度分解,得到红外图像的低频和高频系数,对该系数进行处理,然后反变换得到增强后的红外图像。实验结果表明,该方法在红外图像分割中取得了很好的效果。
关键词:图像处理;非下采样 Contourlet 变换;随机游走;红外图像分割
DOI:10.11907/rjdk.172685
中图分类号:TP317.4
文献标识码:A 文章编号:1672-7800(2018)004-0209-03
Abstract:Infrared images have the characteristics of uneven gray levels, low resolution, and many noises. The traditional image segmentation method is difficult to segment the infrared images effectively. In order to solve this problem, a random walk infrared image segmentation algorithm based on Nonsubsampled Contourlet Transform (NSCT) is designed in this paper. The traditional random walk image segmentation method, the non-uniformity of the gray part of the image can change the path of the random walk algorithm, and the contour of the segmented object is easily affected by the background noise of the image, so we can enhance the image's target contour to suppress the image noise and achieve a better segmentation effect. NSCT transform is a very effective image enhancement method. Firstly, the NSCT transform is used to decompose the infrared image in multi-direction and multi-scale, and the low frequency coefficient and high frequency coefficient of the infrared image can be obtained. The high frequency and low frequency coefficients of the image are processed, and then the enhanced infrared image is changed inversely. The experimental results show that this method has achieved good results in infrared image segmentation.
Key Words:image processing; NSCT; random walk; infrared image segmentation
0 引言
隨着图像处理、分析技术的发展,图像技术已广泛应用于现实生活中,比如人脸识别技术应用于智能手机解锁及支付功能等。图像分割在图像识别和图像分析中占有重要地位,是由图像处理到图像识别的重要步骤。好的图像分割算法可以使图像分析与识别算法更好地运行,因为图像分割的好坏与区域界限准确度都会对后续分析产生影响[1]。红外图像分割是图像分割的一个重要分支,在复杂场景和暗光环境下,相比于可见光图像,红外图像能更准确地表现出场景中的有用信息,所以对红外图像的分割与识别技术也极其重要。……
