A unified polyphase representation of analysis and synthesis filter banks is introduced in this paper, and then the efficient implementation on digital signal processors (DSP) is investigated. Especially, the number of memory accesses, power consumption, processing accuracy and the required instruction cycles are discussed. Firstly, a unified representation is given, and then two types of procedures, SIMO system-based and MISO system-based procedures, are shown, where SIMO and MISO are abbreviations for single-input/multiple-output and multiple-input/single-output, respectively. These procedures are compared to each other. It is shown that the number of data load in SIMO system-based procedure is a half of that in MISO system-based procedure for two-channel filter banks. The implementation of M-channel filter banks is also discussed.

This paper proposes iterative demodulation/decoding for parallel combinatorial spread spectrum (PC/SS) systems. A PC/SS system conveys information data by a combination of pre-assigned orthogonal spreading sequences with polarity. In this paper, convolutional coding with a uniform random interleaver is implemented in channel coding, just like as a serial concatenated coding. A 'soft-in/soft-out' PC/SS demodulator based on a posteriori probability algorithm is proposed to perform the iterative demodulation and decoding. Simulation results demonstrate that the proposed iterative demodulation/decoding scheme bring significant improvement in bit error rate performance. This proposed decoding scheme achieves high-speed transmission by two approaches. One is a puncturing operation, and the other is to increase the number of transmitting sequences. In the latter approach, lower error rate performance is achieved comparing with that the punctured convolutional code is used to increase the information bit rate.

Upper bounds on the bit error rate (BER) for maximum likelihood (ML) decoding are derived in convolutional coded parallel combinatorial spread spectrum (PC/SS) systems over additive white Gaussian noise (AWGN) channels. PC/SS systems can achieve higher data transmission than conventional multicode SS systems. To make the derivation tractable, we put a uniform interleaver between a convolutional encoder and a PC/SS transmitter. Since the PC/SS transmitter is employed as the "inner encoder," the bounds are obtained in a similar manner of the derivation in serially concatenated codes through a uniform interleaver. Two different error patterns in the PC/SS system are considered in the performance analysis. Numerical results show that the derived BER bounds are sufficiently accurate. It is found that the coded PC/SS systems outperform coded all-code-parallel DS/SS systems under the same data rate conditions if the number of pre-assigned PN codes increases.

Several lifting implementation techniques for invertible deniterlacing are proposed in this paper. Firstly, the invertible deinterlacing is reviewed, and an efficient implementation is presented. Next, two deinterlacer-embedded lifting architectures of discrete wavelet transforms (DWT) is proposed. Performances are compared among several architectures of deinterlacing with DWT. The performance evaluation includes dual-multiplier and single-multiplier architectures. The number of equivalent gates shows that the deinterlacing-embedded architectures require less resources than the separate implementaion. Our experimental evaluation of the dual-multiplier architecture results in 0.8% increase in the gate count, whereas the separate implementation of deinterlacing and DWT requires 6.1% increase from the normal DWT architecture. For the proposed single-multiplier architecture, the gate count is shown to result in 4.5% increase, while the separate counterpart yields 10.7% increase.

In this paper, the co-existence of TDMA and W-CDMA spectrum sharing system (TDMA/W-CDMA overlaid system) with cellular architecture is discussed. In this system, both systems share the same frequency band to improve the spectrum efficiency. Overall rate, bit error ratio (BER) and spectrum efficiency of the system are calculated for the forward link (down-link) in the presence of AWGN channel. Taking into account the path loss and shadow fading loss in this system with cellular architecture, W-CDMA applying interference cancellation (IC) shows a substantial difference in spectrum efficiency, the overlaid system can provide a greater overall rate and higher spectrum efficiency than a single multiple access-based system such as TDMA system or W-CDMA system. The interference cancellation can significantly improve BER of the spectrum overlaid system.

The error diffusion filter in this paper is optimized with respect to the ideal blue noise pattern corresponding to a single tone level. The filter coefficients are optimized by the minimization of the squared error norm between the Fourier power spectra of the resulting halftone and the blue noise pattern. During the process of optimization, the binary pattern power spectrum matching algorithm is applied with the aid of a new blue noise model. The number of the optimum filters is equal to that of different tones. The visual fidelity of the bilevel halftones generated by the error diffusion filters is evaluated in terms of a weighted signal-to-noise ratio, Fourier power spectra, and others. Experimental results have demonstrated that the proposed filter set generates satisfactory bilevel halftones of grayscale images.

This letter discusses the performance of online SNR estimation including fading parameter estimation for parallel combinatorial SS (PC/SS) systems. The PC/SS systems are partial-code-parallel multicode SS systems, which have high-rate data transmission capability. Nakagami-m distribution is assumed as fading channel model to cover a wide range of fading conditions. The SNR and fading parameter estimation considered in this letter is based on only a statistical ratio of correlator outputs at the receiver. Numerical results show that SNR estimation performance with fading parameter estimation is close to the one in the case of perfect fading parameter information, if the number of transmitting PN codes is less than a half of assigned PN codes.

Performance of selective Rake (SRake) receiver is evaluated for direct sequence ultra wideband (DS-UWB) communications considering an independent Rayleigh channel having exponentially decaying power delay profile (PDP). BEP performances are shown. The results obtained are compared with similar results in a channel having flat PDP. Assumption of a flat PDP is found to predict the optimum spreading bandwidth to be lower and sub-optimum operating performance beyond optimum spreading bandwidth to be severely worse than that is achievable in a channel having exponentially decaying PDP by employing an SRake receiver having fixed number of combined paths. Optimum spreading bandwidth for SRake in a channel having exponentially decaying PDP is shown to be much larger than the one in a channel having flat PDP; that is specifically a good-news for UWB communications. Effects of partial band interference are also investigated. Interference is found to be less effective in exponentially decaying PDP.

This paper presents a color demosaicing method by introducing iterative asymmetric average interpolation. Missing primary colors on a Bayer pattern color filter array (CFA) are estimated by an asymmetric average interpolation where less intensity variation is assumed to be of stronger significance, before sharpness of an initial estimate is further improved by an iterative procedure. The iteration is implemented by an observation process followed by a restoration process. The former is modeled by blurring followed by CFA sampling and the latter is completely as same as the color demosaicing method initially applied. Experimental results have shown a favorable performance in terms of PSNR and visual appearance, in particular, in sharpness recovery.

In this paper, a coefficient-parameter reduction method is proposed for invertible deinterlacing with variable coefficients. Invertible deinterlacing, which the authors have developed before, can be used as a preprocess of frame-based motion picture codec, such as Motion-JPEG2000 (MJP2), for interlaced videos. When the conventional field-interleaving is used instead, comb-tooth artifacts appear around edges of moving objects. On the other hand, the invertible deinterlacing technique allows us to suppress the comb-tooth artifacts and also to recover an original picture on demand. As previous works, the authors have developed a variable coefficient scheme with a motion detection filter, which realizes adaptability to local characteristics of given pictures. When applying this deinterlacing technique to an image codec, it is required to send coefficient parameters to receivers for original picture recovery. This work proposes a parameter decimation technique and shows that this reduction approach can be achieved without significant loss of comb-tooth suppression capability and improves the quality at high bit-rate decoding.

This paper presents a method of scalable lossless image compression by means of lossy coding. A progressive decoding capability and a full decoding for the lossless rendition are equipped with the losslessly encoded bit stream. Embedded coding is applied to large-amplitude coefficients in a wavelet transform domain. The other wavelet coefficients are encoded by a context-based entropy coding. The proposed method slightly outperforms JPEG-LS in lossless compression. Its rate-distortion performance with respect to progressive decoding is close to that of JPEG2000. The spatial scalability with respect to resolution is also available.

In this paper, a two-dimensional (2-D) binary-valued (BV) lapped transform (LT) is proposed. The proposed LT has basis images which take only BV elements and satisfies the axial-symmetric (AS) property. In one dimension, there is no 2-point LT with the symmetric basis vectors, and the property is achieved only with the non-overlapping basis which the Hadamard transform (HT) has. Hence, in two dimension, there is no 22-point separable ASLT, and only 2-D HT can be the 22-point separable AS orthogonal transform. By taking non-separable BV basis images, this paper shows that a 22-point ASLT can be obtained. Since the proposed LT is similar to HT, it is referred to as the lapped Hadamard transform (LHT). LHT of larger size is shown to be provided with a tree structure. In addition, LHT is shown to be efficiently implemented by a lattice structure.

The checkerboard effect is caused by the periodic time-variant property of multirate filters which consist of up-samplers and digital filters. Although the conditions for some one-dimensional (1D) multirate systems to avoid the checkerboard effect have been shown, the conditions for Multidimensional (MD) multirate systems have not been considered. In this paper, some theorems about the conditions for MD multirate filters without checkerboard effect are derived. In addition, we also consider MD multirate filter banks without checkerboard effect. Simulation examples show that the checkerboard effect can be avoided by using the proposed conditions.

Block delay caused by using fast Fourier transform (FFT), and computational complexity in sampling rate conversion system are considered in this paper. The relationship between the number of block delays and the computational complexity is investigated. The proposed method can avoid the redundant operations of sampling rate conversion completely and moreover provide a good trade-off between the number of block delays and the computational complexity. As a result, ti is shown that with the proposed method, the sampling rate conversion can be realized more efficiently under a small number of block delays.

In this paper, a coefficient-parameter embedding method into Motion-JPEG2000 (MJP2) is proposed for invertible deinterlacing with variable coefficients. Invertible deinterlacing, which the authors have developed before, can be used as a preprocess of frame-based motion picture codec, such as MJP2, for interlaced videos. When the conventional field-interleaving is used instead, comb-tooth artifacts appear around edges of moving objects. On the other hand, the invertible deinterlacing technique allows us to suppress the comb-tooth artifacts and also guaranties recovery of original pictures. As previous works, the authors have developed a variable coefficient scheme with a motion detector, which realizes adaptability to local characteristics of given pictures. However, when this deinterlacing technique is applied to a video codec, coefficient parameters have to be sent to receivers for original picture recovery. This paper proposes a parameter-embedding technique in MJP2 and constructs a standard stream which consists both of picture data and the parameters. The parameters are embedded into the LH1 component of wavelet transform domain through the ROI (region of interest) function of JPEG2000 without significant loss in the performance of comb-tooth suppression. Some experimental results show the feasibility of our proposed scheme.

This paper proposes a Poisson denoising method with a union of directional lapped orthogonal transforms (DirLOTs). DirLOTs are 2-D non-separable lapped orthogonal transforms with directional characteristics under the fixed-critically-subsampling, overlapping, orthonormal, symmetric, real-valued and compact-support property. In this work, DirLOTs are used to generate symmetric orthogonal discrete wavelet transforms and then a redundant dictionary as a union of unitary transforms. The multiple directional property is suitable for representing natural images which contain diagonal textures and edges. Multiple DirLOTs can overcome a disadvantage of separable wavelets in representing diagonal components. In addition to this feature, multiple DirLOTs make transform-based denoising performance better through the redundant representation. Experimental results show that the combination of the variance stabilizing transformation (VST), Stein's unbiased risk estimator-linear expansion of threshold (SURE-LET) approach and multiple DirLOTs is able to significantly improve the denoising performance.

In this paper, we propose two parallel processing methods for multidimensional (MD) sampling lattice alteration. The use of our proposed methods enables us to alter the sampling lattice of a given MD signal sequence in parallel without any redundancy caused by up- and down-sampling, even if the alteration is rational and non-separable. Our proposed methods are provided by extending two conventional block processing techniques for FIR filtering: the overlap-add method and the overlap-save method. In these proposed methods, firstly a given signal sequence is segmented into some blocks, secondly sampling lattice alteration is implemented for each block data individually, and finally the results are fitted together to obtain the output sequence which is identical to the sequence obtained from the direct sampling lattice alteration. Besides, we provide their efficient implementation: the DFT-domain approach, and give some comments on the computational complexity in order to show the effectiveness of our proposed methods.

This work proposes an exponential computation with low-computational complexity and applies this technique to the expectation-maximization (EM) algorithm for Gaussian mixture model (GMM). For certain machine-learning techniques, such as the EM algorithm for the GMM, fast and low-cost implementations are preferred over high precision ones. Since the exponential function is frequently used in machine-learning algorithms, this work proposes reducing computational complexity by transforming the function into powers of two and introducing a look-up table. Moreover, to improve efficiency the look-up table is scaled. To verify the validity of the proposed technique, this work obtains simulation results for the EM algorithm used for parameter estimation and evaluates the performances of the results in terms of the mean absolute error and computational time. This work compares our proposed method against the Taylor expansion and the exp( ) function in a standard C library, and shows that the computational time of the EM algorithm is reduced while maintaining comparable precision in the estimation results.

This letter proposes an image fusion method which adopts a union of multiple directional lapped orthogonal transforms (DirLOTs). DirLOTs are used to generate symmetric orthogonal discrete wavelet transforms and then to construct a union of unitary transforms as a redundant dictionary with a multiple directional property. The multiple DirLOTs can overcome a disadvantage of separable wavelets to represent images which contain slant textures and edges. We analyse the characteristic of local luminance contrast, and propose a fusion rule based on interscale relation of wavelet coefficients. Relying on the above, a novel image fusion method is proposed. Some experimental results show that the proposed method is able to significantly improve the fusion performance from those with the conventional discrete wavelet transforms.

In this paper, theoretical properties of deinterlacer banks are analyzed. Deinterlacer banks are novel filter banks in the sense that a progressive video sequence is separated into two progressive video sequences of a half frame rate and, furthermore, interlaced sequences are produced as intermediate data. Unlike the conventional filter banks, our deinterlacer banks are constructed in a way unique to multidimensional systems by using invertible deinterlacers, which the authors have proposed before. The system is a kind of shift-varying filter banks and it was impossible to derive the optimal bit-allocation control without any equivalent parallel filter banks. This paper derives an equivalent polyphase matrix representation of the whole system and its equivalent parallel structure, and then shows the optimal rate allocation for the deinterlacer banks. Some experimental results justify the effectiveness of the optimal rate allocation through our theoretical analysis.