A Region-Based Scheme Using RKLT and Predictive Classified Vector Quantization

doi:10.1006/cviu.1999.0776

Computer Vision and Image Understanding

Volume 75, Issue 3, September 1999, Pages 269-280

https://doi.org/10.1006/cviu.1999.0776 Get rights and content

Abstract

This paper proposes a compression scheme for face profile images based on three stages, modelling, transformation, and the partially predictive classified vector quantization (CVQ) stage. The modelling stage employs deformable templates in the localisation of salient features of face images and in the normalization of the image content. The second stage uses a dictionary of feature-bases trained for profile face images to diagonalize the image blocks. At this stage, all normalized training and test images are spatially clustered (objectively) into four subregions according to their energy content, and the residuals of the most important clusters are further clustered (subjectively) in the spectral domain, to exploit spectral redundancies. The feature-basis functions are established with the region-based Karhunen–Loeve transform (RKLT) of clustered image blocks. Each image block is matched with a representative of near-best basis functions. A predictive approach is employed for mid-energy clusters, in both stages of search for a basis and for a codeword from the range of its cluster. The proposed scheme employs one stage of a cascaded region-based KLT-SVD and CVQ complex, followed by residual VQ stages for subjectively important regions. The first dictionary of feature-bases is dedicated to the main content of the image and the second is dedicated to the residuals. The proposed scheme is experimented in a set of human face images.

References (41)

K. Aizawa et al.
Model-based analysis-synthesis image coding (MBASIC) system for a person's face
Signal Proces. Image Commun.
(1989)
R.D. Dony et al.
Compression of SAR images using KLT, VQ, and mixture of principle components
IEEE Pro. Radar, Sonar Navig.
(1997)
G. Fernandez and C. M. Wittenbrink, Region based KLT for multispectral image compression, in Proc. of EUSIPCO'96, Sept....
R. W. Brause, Cascaded vector quantization by non-linear PCA network layers, in Proc. IEEE 6. Int. Conf. on Tools with...
T. Y. Kim, B. H. Kim, S. O. Kim, and H. M. Choi, KLT-based vector quantization using PCNN, in Proc. IEEE, Int. Conf. on...
M. Turk et al.
Eigenfaces for recognition
J. Cogn. Neuros.
(1991)
R. D. Dony and S. HaykinOptimally integrated adaptive learning, in IEEE, Proc. ICASSP, June 1993, Vol. 1,...
B. R. Epstein, R. Hingorani, J. M. Shapiro, and M. Czigler. Multispectral KLT-Wavelet Data Compression for Landsat...
P. Waldemar and T. A. Ramstad, Hybrid KLT-SVD image compression, in IEEE Proc. ICASSP, Apr. 1997, Vol. , pp....
H. Shen and R. L. Baker, A finite state/frame difference interpolative vector quantizer for low rate image sequence...

T.D. Lookabaugh et al.

High-resolution quantization theory and the vector quantization advantage

IEEE Trans. Inf. Theory

(1989)

A. Hoang et al.

Block quantization of correlated Gaussian random variables

IEEE Trans. Commun. Systems

(1963)

N.S. Jayant et al.

Digital Coding of Waveforms

(1984)

M. Rabbani et al.

Digital Image Compression Techniques

(1991)

M. Vetterli

Wavelets and Subband Coding

(1995)

A.N. Akansu et al.

Multiresolution Signal Decomposition

(1992)

A. Gersho et al.

Vector Quantization and Signal Compression

(1992)

D.E. Pearson

Developments in model-based video coding

Proc. IEEE

(1995)

C. Zetzsche, E. Barth, and B. Wegmann, The importance of intrinsically two-dimensional image features in biological...

W.K. Pratt

Digital Image Processing

(1991)

Cited by (8)

Compression of aerodynamic databases using high-order singular value decomposition
2010, Aerospace Science and Technology
Citation Excerpt :
Also, Kanth et al. [12] have developed a novel SVD technique that allows for image reconstruction with an error caused by the approximate computation smaller than 10%. Another SVD based methodology for face image recognition has been reported by Sakalli et al. [13], who developed a step-wise approach to the problem that involved the Karhunen–Loeve transform of clustered image blocks. More recently, a variant of SVD called partial singular value decomposition has been applied by Tougas and Spiteri [16] to latent semantic indexing, which is an information retrieval method.
A methodology based on high-order singular value decomposition is presented to compress multidimensional (with the various dimensions associated with both the spatial coordinates and parameter values) aerodynamic databases. The method is illustrated with a database containing computational fluid dynamics calculations of the outer flow around a wing, with two free parameters, the Mach number and the angle of attack. Comparison is made between the results of compressing just one flow snapshot (for fixed values of the parameters), compressing a one-parameter family of snapshots, and compressing the whole database. Several compressing strategies are also discussed that deal with (a) treating the flow variables separately or considering all flow variables at a time, (b) considering the whole flow domain simultaneously or dividing it into blocks, and (c) using various measures of errors. The main conclusion is that a large compression factor is generally obtained. Furthermore, the compression factor increases exponentially as the dimension of the database increases for any fixed error, namely the compression factor increases by an order of magnitude with each new database dimension for an error level of 1%.
Compression of facial images using the K-SVD algorithm
2008, Journal of Visual Communication and Image Representation
The use of sparse representations in signal and image processing is gradually increasing in the past several years. Obtaining an overcomplete dictionary from a set of signals allows us to represent them as a sparse linear combination of dictionary atoms. Pursuit algorithms are then used for signal decomposition. A recent work introduced the K-SVD algorithm, which is a novel method for training overcomplete dictionaries that lead to sparse signal representation. In this work we propose a new method for compressing facial images, based on the K-SVD algorithm. We train K-SVD dictionaries for predefined image patches, and compress each new image according to these dictionaries. The encoding is based on sparse coding of each image patch using the relevant trained dictionary, and the decoding is a simple reconstruction of the patches by linear combination of atoms. An essential pre-process stage for this method is an image alignment procedure, where several facial features are detected and geometrically warped into a canonical spatial location. We present this new method, analyze its results and compare it to several competing compression techniques.
Passport photo compression technique with JPEG2000
2013, 2013 IEEE International Conference on Mechatronics and Automation, IEEE ICMA 2013
General scalable image compression for formal ID picture
2012, International Journal of Digital Content Technology and its Applications
Sparse and redundant representations: From theory to applications in signal and image processing
2010, Sparse and Redundant Representations: From Theory to Applications in Signal and Image Processing
Formal photograph compression in JPEG2000 coding
2010, Proceedings - 2010 International Symposium on Intelligence Information Processing and Trusted Computing, IPTC 2010

View all citing articles on Scopus

^f1: [email protected]

View full text

Regular ArticleA Region-Based Scheme Using RKLT and Predictive Classified Vector Quantization

Abstract

Signal Proces. Image Commun.

Compression of SAR images using KLT, VQ, and mixture of principle components

IEEE Pro. Radar, Sonar Navig.

Eigenfaces for recognition

J. Cogn. Neuros.