基于形状和纹理的图像检索

Shape-texture Based Image Retrieval

构造了Radon变换的不变量,提出一种新的基于形状和纹理的图像检索方法。对小波图像边缘提取Radon不变量作为形状特征,同时提取小波各频道能量作为纹理特征。然后将形状特征和纹理特征分别进行高斯归一化,计算图像形状和纹理的相似度。最后,利用形状和纹理相似度的加权和进行图像检索。试验结果表明该方法对噪声具有较强的鲁棒性,具有尺度、平移和旋转不变性。

In this paper, the invariant using Radon transform is constructed and a new image retrieval algorithm based on shape and texture information is presented. It turns to be getting multi-scale edge images by the wavelet modulus maximum after decomposing the images with wavelet transform in the algorithm. The Radon invariant is extracted to be the shape feature of image to the multi- scale edge images and the energy of every sub frequency bands is extracted to be the texture feature of image. The Gaussian model is used to normalize the different sub-characters distance to the shape feature of image and the texture feature of image separately. The shape similarity of images and the texture similarity of images are computed by the Euclidean distance separately. Finally, the weigh- ted sums of the shape similarity of images and the texture similarity of images can be the similarity between the querying image and other images. Experiments indicate that this method is robustness to the noises in image＇s similarity retrieval and has the property of shift, scale and rotation invariance.