Title :
Classification oriented embedded image coding
Author_Institution :
Dept. of Electr. & Comput. Eng., Duke Univ., Durham, NC, USA
Abstract :
This paper discusses the efficient compression of images to improve the classification associated with the decoded imagery. The set partitioning in hierarchical trees (SPIHT) algorithm, an efficient wavelet-based progressive image-compression scheme, was originally designed to minimize the mean-squared error (MSE) between the original and decoded imagery. The image is first segmented at the encoder by an unsupervised method using a hidden Markov tree (HMT) mixture model in the wavelet domain. By using the kernel matching pursuits (KMP) method the recognition importance of each wavelet subband is estimated. By comparison using synthesized data, the compression and classification performance of the modified SPIHT algorithm is comparable to Bayes TSVQ, along with the advantages of fast speed and no requirement of codebook design and possibly transmission.
Keywords :
data encapsulation; hidden Markov models; image classification; image coding; image segmentation; iterative methods; mean square error methods; tree codes; trees (mathematics); vector quantisation; wavelet transforms; Bayes tree structured vector quantization; MSE; classification oriented embedded image coding; codebook design; data synthesis; decoded imagery; hidden Markov tree; image recognition; image segmentation; kernel matching pursuit method; mean-squared error; set partitioning in hierarchical tree algorithm; unsupervised method; wavelet domain; wavelet subband; wavelet-based progressive image-compression scheme; Algorithm design and analysis; Bit rate; Decoding; Hidden Markov models; Image coding; Image recognition; Image segmentation; Matching pursuit algorithms; Partitioning algorithms; Wavelet coefficients;
Conference_Titel :
Data Compression Conference, 2004. Proceedings. DCC 2004
Print_ISBN :
0-7695-2082-0
DOI :
10.1109/DCC.2004.1281506
