Traditional design methods of two-dimensional large-scale filter banks suffer from high-complexity. This paper presents an algorithm to design two-dimensional double-prototype fully oversampled Discrete Fourier Transform (DFT) modulated filter bank with Nearly Perfect Reconstruction (NPR). The algorithm is based on bi-iterative scheme, where the design issue is formulated into an unconstrained optimization issue whose objective function is the weighted sum of the transfer distortion and the aliasing distortion of the filter bank, and the stopband energy of the Prototype Filters (PFs). By exploiting the gradient information, the optimization problem can be efficiently solved by utilizing the bi-iterative scheme. The matrix inverse identity and the fast algorithm for Toeplitz-block Toeplitz matrix inversion are employed to dramatically reduce the computational cost of the iterative procedure. The theoretical analysis and numerical experiments are carried out to show that compared with the existing methods, the new algorithm possesses much lower computational cost and can be used to design large-scale two-dimensional filter bank with better overall performance.
VAIDYANATHAN P P. Multirate Systems and Flter Banks[M]. Englewoo Cliffs: N.J.,Prentice Hall, 1993: 188-272.
[2]
LIN Y P and VAIDYANATHAN P P. Theory and design of two-dimensional filter bank: A review[J]. Multidimensional System & Signal Processing, 1996, 7(3-4): 263-330. doi: 10.1007/BF01826246.
[3]
GAWANDE J P, RAHULKAR A D, and HOLAMBE R S. Design of new class of regular biorthogonal wavelet filter banks using generalized and hybrid lifting structures[J]. Signal Image and Video Processing, 2015, 9(1): 265-273. doi: 10.1007/s11760-015-0814-0.
[4]
SHUI Penglang. Image denoising using 2-D separable oversampled DFT modulated filter banks[J]. IET Image Processing, 2009, 3(3): 163-173. doi: 10.1049/iet-ipr.2007.0218.
[5]
SUZUKI T and KUDO H. Two-dimensional non-separable block-lifting-based M-channel biorthogonal filter banks[C]. European Signal Processing Conference, Lisbon, 2014: 291-295.
[6]
RAJAPAKAHA N, MADANAYAKE A, and BRUTON LT. 2D space-time wave-digital multi-fan filter banks for signals consisting of multiple plane waves[J]. Multidimensional Systems and Signal Processing, 2014, 25(1): 17-39. doi: 10.1007/s11045-012-0183-6.
[7]
SUZUKIT and KUDO H. Two-dimensional non-separable block-lifting structure and its application to M-channel perfect reconstruction filter banks for lossy-to-lossless image coding[J]. IEEE Transactions on Image Processing, 2015, 24(12): 4943-4951. doi: 10.1109/TIP.2015.2472294.
[8]
WILBUR M R, DAVIDSON T N, and REILLY J P. Efficient design of oversampled NPR GDFT filter banks[J]. IEEE Transactions on Signal Processing, 2004, 52(7): 1947-1963. doi: 10.1109/TSP.2004.828936.
[9]
SHUI Penglang and JIANG Junzheng. Two-dimensional 2× oversampled DFT modulated filter banks and critically sampled modified DFT modulated filter banks[J]. IEEE Transactions on Signal Processing, 2010, 58(11): 5597-5611. doi: 10.1109/TSP.2010.2059016.
[10]
JIANG Junzheng and ZHOU Fang. Iterative design of two-dimensional critically sampled MDFT modulated filter banks[J]. Signal Processing, 2013, 93(11): 3124-3132. doi: 10.1016/j.sigpro.2013.03.022.
[11]
JIANG Junzheng, ZHOU Fang, SHUI Penglang, et al. Theory and design of two-dimensional DFT modulated filter bank with arbitrary modulation and decimation matrices[J]. Digital Signal Processing, 2015, 44(1): 123-130. doi: 10.1016/ j.dsp.2015.05.012.
[12]
JIANG Junzheng and SHUI Penglang. Design of 2D linear phase DFT modulated filter banks using bi-iterative second-order cone program[J]. Signal Processing, 2010, 90(12): 3065-3077. doi: 10.1016/j.sigpro.2010.05.011.
[13]
JIANG Junzheng and SHUI Penglang. Design of 2D oversampled linear phase DFT modulated filter banks via modified Newton’s method[J]. Signal Processing, 2012, 92(6): 1411-1421. doi: 10.1016/j.sigpro.2011.11.029.
[14]
JIANG Junzheng, ZHOU Fang, and OUYANG Shan. Design of two-dimensional large-scale DFT modulated filter banks[J]. IET Signal Processing, 2013, 7(9): 807-813. doi: 10.1049/ iet-spr.2012.0327.
[15]
ZHOU Fang, JIANG Junzheng, and SHUI Penglang. Fast design of 2D fully oversampled DFT modulated filter bank using Toeplitz-block Toeplitz matrix inversion[J]. Signal Processing, 2015, 111: 194-198. doi: 10.1016/j.sigpro.2014.12.021.
JIANG Junzheng, WANG Xiaolong, and SHUI Penglang. Novel method for designing DFT modulated filter banks[J]. Journal of Xidian University, 2010, 37(4): 689-693. doi: 10.3969/j.issn.1001-2400.2010.04.019.
[17]
PETERSEN K B and PETERSEN M S. The Matrix Cookbook[OL]. http://www2.imm.dtu.dk/pubdb/p.php, 2012.11.
[18]
WAX M and KAILATH T. Efficient inversion of Toeplitz-block Toeplitz matrix[J]. IEEE Transactions on Acoustics, 1983, 31(5): 1218-1221. doi: 10.1109/TASSP.1983.1164208.
