一种基于小波纹理特征的目标检测方法

A Target Detection Approach Based on Wavelet Texture Characterization

给出了一种利用小波分块分解和灰度共生矩阵特征来提取目标的方法。首先对图像进行分块小波变换,然后求分块灰度共生矩阵并且计算小波共生矩阵特征向量,选取纹理特征最大的作为种子区域;最后利用均值聚类的方法进行目标标记。实验结果证明,能够检测红外和可见光图像中各种类型的目标。

An approach of detecting targets based on wavelet transform and gray level co-occurrence matrix was presented. Wavelet transform is used to decompose an image into blocks; the gray level co-occurrence matrix and each block wavelet co-occurrence feature value are evaluated. A block with max texture value is selected as a seed block and mean clustering method is used to mark blocks. Experiments show this method can detect various targets in IR and visible images.