小波变换在图像匹配技术中的应用研究

Application of Wavelet Transform in Image Matching Technology

针对单一图像匹配算法抗噪声能力和抗几何形变能力差、以及计算量较大等问题,提出一种基于小波变换后图像统计特征的图像匹配识别算法.利用小波对图像进行多个尺度的二维小波变换,分析反映图像在不同尺度和不同方向上能量分布情况,通过计算统计特征量和标准图库中图像统计特征量之间的夹角,确定待识别目标,利用该方法对图书条形码进行识别.研究结果表明:采用小波变时频特性对图像进行去噪处理以及边缘检测,能够剔除图像噪声干扰并充分保留原始图像信息;将小波变换后的近似系数、水平、垂直和对角部分的细节系数作为统计特征量,提取的特征量跟模板图像的统计特征量进行相似度匹配,将两个特征向量的夹角余弦作为相似度的衡量指标,根据夹角余弦的大小确定该区域是否目标区域.

The anti-noise and resistance to geometric deformation ability of a single image matching algorithm is poor .When the color difference between the image background and target area is very small ,the matching algorithm based on gray can not effectively distinguish between background area and the target area .Above all ,its calculation amount is large and it has lower accuracy .Therefore ,a kind of image matching recognition algorithm based on statistical characteristics of the image after the wavelet transform is proposed .This algorithm adopts wavelet to perform multiple scale two-dimensional wavelet transform for the image .The corresponding progeny image energy distribution in different direction and different scale will be taken as statistical characteristics .The targets to be recognized are identified through the angle＇s cosine value between the extracted statistical features from the targets and the statistical characteristics of the standard library image .Finally ,the bar codes of books are adopted to verify this method and the results show that the algorithm has good recognition effect for the different characters in bar codes .