一种新的基于对称色彩空域特征的图像匹配方法

Novel Image Matching Algorithm Based on Symmetrical Color-Spatial Feature

在目前的基于内容的图像检索中 ,颜色是使用最广泛的进行图像特征匹配的特征之一 .在基于颜色的图像特征匹配方法中引入空域信息对于确保匹配准确率是十分必要的 .颜色特征的优点在于对尺度、旋转的不变性 ,而引入空域信息后 ,会消除这种好的性质 .因而需要解决如何在引入空域信息的同时 ,兼顾不变性的问题 .提出一种基于对称色彩空域特征的图像匹配方法 .该方法在保留了颜色特征不变性的基础上 ,通过引入对称的空域信息 ,既提高了图像特征匹配的准确性 ,又消除了图像变形对图像特征匹配的影响 .最后的试验表明我们的算法在一定程度上解决了图像变形 ,特别是对称变形对图像匹配的影响 .

Color is the most common used feature for image matching in CBIR (Content Based Image Retrieval ) system by now. It is ve ry necessary to provide spatial information of color for increasing the accuracy of image matching based on color feature. The advantage of color feature is ind ependent of image distortion such as resizing and rotating. However after combin ing with spatial information, this advantage would be disappear. This paper prop osed a novel image-matching algorithm that based on symmetrical color-sp atial feature. By using the symmetrical spatial information for color feature, t his algorithm reserved the independence of color feature and eliminates the infl uence of distortion to the image matching while increasing the accuracy. Exp eriments show that the new algorithm reduces the mismatch caused by image distor tion especially the symmetrical distortion.