提出了一种基于小波分解的具有旋转不变特性的三级特征图像快速匹配方法,利用小波分解的高频分量得到图像边缘,进而提取具有旋转不变特性的边缘特征;小波分解的低频分量则用于提取同样具有旋转不变性的图像尺度共生矩阵特征.前一个特征反映了图像边缘间的空间关系,后者则体现了图像灰度间的空间关系,利用这两类特征构成一个三级决策树对图像进行搜索,以查找搜索图中是否包含样本图,对满足匹配条件的若干候选子图,给出具有最大相似特性的一个,实验结果表明,较之于传统的模板匹配方法,该方法具有快速,高精度,抗干扰能力强等优点.

Image matching is very important in image processing, especially in remote sensing image processing. A nice matchimg method must possess the advantages of high matching precision, fast matching speed and the property of rotation invariant. In this paper, we proposed a novel method for fast image matching based on wavelet analysis with rotation invariant property. The edge of processing image is extracted from the detailed image of wavelet decomposition, which is used to calculate the distance moments of the image; the blurred images are designed for extracting the scale based Concurrent Matrix (SCM) features. A decision tree based on these two types of feature is then formed to implement the matching procedure. Experiments also show that this method is faster, more accurate and better noise robust compared to the traditional model based methods.