Image Feature Description by Frequent Patterns

The classification of image data becomes important, due to the increasing application of digital images, unsupervised classification technology with high capacity is necessary for processing digital images. In this paper, we propose an unsupervised approach of image pattern description and classification. In order to collect frequently appeared patterns in images, a compressibility feature space is built in an unsupervised manner. Based on this feature space the proposed approach transforms images to sequences, which are then divided into segments and replaced by characters. Finally, the similarities among compressibility vectors of texts are used for classification, instead of using texts themselves. Our experiments showed that the proposed approach is effective.

[1]  Matteo Masotti,et al.  A ranklet-based image representation for mass classification in digital mammograms. , 2006, Medical physics.

[2]  Aidong Zhang,et al.  Image Decomposition and Representation in Large Image Database Systems , 1997, J. Vis. Commun. Image Represent..

[3]  Zhouyu Fu,et al.  Recognition of Pornographic Web Pages by Classifying Texts and Images , 2007, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[4]  Alan Wee-Chung Liew,et al.  Information-Based Color Feature Representation for Image Classification , 2007, 2007 IEEE International Conference on Image Processing.

[5]  Abraham Lempel,et al.  Compression of individual sequences via variable-rate coding , 1978, IEEE Trans. Inf. Theory.

[6]  Jiawei Han,et al.  Image clustering with tensor representation , 2005, ACM Multimedia.

[7]  David G. Stork,et al.  Pattern classification, 2nd Edition , 2000 .

[8]  Shigeo Abe DrEng Pattern Classification , 2001, Springer London.

[9]  Ken Sugawara,et al.  A New Pattern Representation Scheme Using Data Compression , 2002, IEEE Trans. Pattern Anal. Mach. Intell..