Effective Wavelet Information Distribution Entropy Techniques for Image Retrieval

曾智勇,刘侍刚
2010-01-01
Abstract:Having analyzed the drawbacks of image index techniques based on wavelet domain,a novel image retrieval algorithm based on wavelet information distributed entropy is proposed in this paper.First,an image is segmented several sub-images by this algorithm.Second,each sub-image is decomposed three stages using wavelets,and each sub-band image by wavelet transform is processed to obtain a texture channel,so the computatinal complexity of this method is greatly reduced.Third,the wavelet information distributed entropy of each sub-image is computed using wavelet texture histogram for probability density funcion,therefore,not only is the image feature presented compactly,but also image retrieval speed is greatly quicken.Experimental results show that the proposed algorithm is effective for image retrieval.
What problem does this paper attempt to address?