Image retrieval based on salient points and keyblocks
Guoyu Wang · Journal of Computer Applications · 2006
A novel image retrieval method was proposed based on salient points and keyblocks. First salient points were extracted using wavelet transform, then the image was segmented into equal-size blocks. These blocks could be divided into two classes, one with the salient points and the other not. Low-level features such as color were extracted to describe block. By adjusting the weights of the two classes in image retrieval, different results depending on our interest in local or global similarity could be got.