Adaptive Nonseparable Wavelet Transform via Lifting and its Application to Content-Based Image Retrieval

被引:61
作者
Quellec, Gwenole [1 ,2 ]
Lamard, Mathieu [2 ,3 ]
Cazuguel, Guy [1 ,2 ]
Cochener, Beatrice [2 ,3 ,4 ]
Roux, Christian [1 ,2 ]
机构
[1] UEB, Inst TELECOM, TELECOM Bretagne, Dept ITI, F-29200 Brest, France
[2] ScInBioS, INSERM, U650, IFR 148, F-29200 Brest, France
[3] Univ Bretagne Occidentale, F-29200 Brest, France
[4] CHU Brest, Serv Ophtalmol, F-29200 Brest, France
关键词
CBIR; lifting scheme; multidimensional wavelet; nonseparable wavelet; wavelet adaptation; FEATURES; DESIGN;
D O I
10.1109/TIP.2009.2030479
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We present in this paper a novel way to adapt a multidimensional wavelet filter bank, based on the nonseparable lifting scheme framework, to any specific problem. It allows the design of filter banks with a desired number of degrees of freedom, while controlling the number of vanishing moments of the primal wavelet ((N) over tilde moments) and of the dual wavelet (moments). The prediction and update filters, in the lifting scheme based filter banks, are defined as Neville filters of order (N) over tilde and N, respectively. However, in order to introduce some degrees of freedom in the design, these filters are not defined as the simplest Neville filters. The proposed method is convenient: the same algorithm is used whatever the dimensionality of the signal, and whatever the lattice used. The method is applied to content-based image retrieval (CBIR): an image signature is derived from this new adaptive nonseparable wavelet transform. The method is evaluated on four image databases and compared to a similar CBIR system, based on an adaptive separable wavelet transform. The mean precision at five of the nonseparable wavelet based system is notably higher on three out of the four databases, and comparable on the other one. The proposed method also compares favorably with the dual-tree complex wavelet transform, an overcomplete nonseparable wavelet transform.
引用
收藏
页码:25 / 35
页数:11
相关论文
共 39 条
[11]  
Do MN, 2002, IEEE IMAGE PROC, P357
[12]  
Goldberg DE., 1989, GENETIC ALGORITHMS S, V13
[13]  
Golub G., 1965, Journal of the Society for Industrial and Applied Mathematics, Series B: Numerical Analysis, V2, P205, DOI [10.1137/0702016, DOI 10.1137/0702016]
[14]   A new method of estimating wavelet with desired features from a given signal [J].
Gupta, A ;
Joshi, SD ;
Prasad, S .
SIGNAL PROCESSING, 2005, 85 (01) :147-161
[15]  
Heath M, 1998, COMP IMAG VIS, V13, P457
[16]   Nonlinear multiresolution signal decomposition schemes - Part II: morphological wavelets [J].
Heijmans, HJAM ;
Goutsias, J .
IEEE TRANSACTIONS ON IMAGE PROCESSING, 2000, 9 (11) :1897-1913
[17]  
KOVACEVIC J, 1990, 1990 IEEE INTERNATIONAL SYMP ON CIRCUITS AND SYSTEMS, VOLS 1-4, P2004, DOI 10.1109/ISCAS.1990.112128
[18]   NONSEPARABLE MULTIDIMENSIONAL PERFECT RECONSTRUCTION FILTER BANKS AND WAVELET BASES FOR RN [J].
KOVACEVIC, J ;
VETTERLI, M .
IEEE TRANSACTIONS ON INFORMATION THEORY, 1992, 38 (02) :533-555
[19]   Wavelet families of increasing order in arbitrary dimensions [J].
Kovacevic, J ;
Sweldens, W .
IEEE TRANSACTIONS ON IMAGE PROCESSING, 2000, 9 (03) :480-496
[20]  
LAMARD M, 2007, 29 ANN INT C IEEE EN