基于颜色空间分布特征的图像检索

An Image Retrieval Method Based on Color-Spatial Feature

目前,基于颜色特征的图像检索大多是以图像的颜色直方图作为颜色特征,这种图像检索方法有简单高效的优点,但丢失了颜色的空间分布信息,该文从CT图像重建的理论中得到启发,将对一幅图像从几个方向的投影图作为这幅图像的颜色特征分布。为进一步减少检索时运算的数据量,对图像做小波分解,然后对分解后图像的低频子带做Radon变换得到颜色空间分布的特征向量,并根据这个特征进行检索,实验表明,当检索图像中有明显的颜色目标时,该方法比传统的颜色直方图法更精确,颜色空间性更强,而且检索用时更短。

As an important visual information,the color histogram is now widely used in image retrieval.Color histogram-based image retrieval method is simple and efficient but losing the spatial distribution information of the color.Enlightened by the method of 2-D image reconstruction,this paper presents a novel and efficient scheme for feature exaction and color image retrieval.The paper also puts forward a method of image's color-spatial feature extraction using Radon transform.In order to reduce the computational complexity,authors make use of wavelet decomposition.At first,the image is decomposed with 2-D Mallat algorithm,and then the low frequency component of each image is processed by Radon transformation to generate the color-spatial feature.In this way,authors get the moment feature matrix of images and the matrix is saved accompanying with the images.In Content Based Image Retrieval(CBIR),the similarity between two images is computed according to their moment feature matrix.The experimental results show that when there are obvious objects in the image,Radon transform method is more accurate and efficient in CBIR than color histogram method.In addition,Radon transform-based image retrieval method costs much less time than color histogram method.