基于纹理一致性测度的汽车车徽分割方法

Segmentation Method for Car Emblem Based on Texture Homogeneity Measure

车徽是汽车的品牌标志,包含了汽车的重要信息。更为重要的是,车徽是难以更换的,可作为汽车分类与识别的重要依据。首先,对原始图像进行小波变换,求取其垂直纹理测度;其次,选取合适的阈值二值化测度图像并进行形态闭运算;然后,根据位置信息进行车徽粗分割;最后,根据纹理一致性测度提取车徽。实验结果表明了提出的车徽分割方法的有效性。

The car emblem is a label of the car and contains important information about the car.Because car emblem exchange is very difficult,it can provide a crucial basis for classifying and identify a car.This method is based on texture homogeneity measure which is defined in the paper.The main steps of the proposed method are:firsthy,perform wavelet transform on the original image and calculate its vertical texture measure;secondly,select a appropriate threshold to segment the measure image and then implement morphological closing operation on the binary image;thirdly,compute the coarse position of the car emblem;in the end,extract the car emblem from the background with homogeneity measure.The experimental results show that the proposed algorithm is robust to segmentation of the car emblem.