基于小波包和数学形态学结合的图像特征提取方法  

Image characteristic extraction method based on wavelet packet and mathematical morphology

作　　者：张立国[1] 杨瑾[1] 李晶[1] 任晓丽[1] 上官寒露[1] 

机构地区：[1]燕山大学电气工程学院测试计量技术及仪器河北省重点实验室,秦皇岛066004

出　　处：《仪器仪表学报》2010年第10期2285-2290,共6页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(61077071;51075349);河北省自然科学基金(E2008000812)资助项目

摘　　要：提出了一种基于小波包和数学形态学相结合的图像特征提取方法,应用于车牌字符等的图像提取及边缘检测。采用了小波包对图像进行分解并重构其近似部分,用形态学膨胀和腐蚀等形态学基本运算以及形态学梯度对图像进行边缘检测,并应用形态学重构填充了车牌字符的空隙,以便后续的字符识别。仿真实验表明与Edge边缘检测算子相比较,该方法能够更好地提取图像特征,检测出的边缘更清晰,并更好地填充了车牌字符的空隙。This paper proposes an image characteristic extraction method based on wavelet packet and mathematical morphology, and applies the method to the character extraction and edge detection of the images of license plate and etc. Wavelet packet is used to decompose the images and reconstruct their approximate part; edge detection of the images is carried out using morphological basic operations, such as dilation, erosion and morphological gradient; and the void of the plate characters is filled using morphological reconstruction for the identification of the characters of the license plate. Simulation experiments show that compared with edge detection operator of ＂edge function＂, the proposed method can extract the characteristic of the images better; and the extracted edge is obviously clear and the void of the characters is filled completely.

关 键 词：数学形态学 小波包 边缘检测 特征提取 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]