A new stereo echo canceler with input slides and counter-lateralization is proposed. Convergence of filter coefficients to the correct echo paths is obtained by pre-processing which delays the input signal periodically by one sample in one of the two channels. The time difference between the two stereo components of the input signals causes a shift of the sound image. This shift is compensated for by presenting the delayed component of the stereo signals to a loudspeaker at a higher intensity, and the other component at a lower intensity. Correct echo-path identification is analytically shown in a more general form than in the preceding literatures. A subjective listening test shows that this method is more effective for vocal musics. The processed signals are scored 0.45 lower than the original input signals, using the ITU-R five-grade impairment scale.

A novel shadow removal approach is proposed by using block-wise transform domain shadow detection. The approach is based on the fact that the spatial frequency distributions on normal background areas and those under casted shadows from foreground objects are the same. The proposed approach is especially useful for silhouette extraction by using the Gaussian Mixture background Model (GMM) foreground segmentation in the transform domain, because the frequency distribution has already been calculated in the foreground segmentation. The stable shadow removal is realized, due to the transform domain implementation.

This paper describes an efficient image enhancement method based on the Multi-Scale Retinex (MSR) approach for pre-processing of video applications. The processing amount is drastically reduced to 4 orders less than that of the original MSR, and 1 order less than the latest fast MSR method. For the efficient processing, our proposed method employs multi-stage and multi-rate filter processing which is constructed by a x-y separable and polyphase structure. In addition, the MSR association is effectively implemented during the above multi-stage processing. The method also modifies a weighting function for enhancement to improve color rendition of bright areas in an image. A variety of evaluation results show that the performance of our simplified method is similar to those of the original MSR, in terms of visual perception, contrast enhancement effects, and hue changes. Moreover, experimental results show that pre-processing of the proposed method contributes to clear foreground object separation.

A novel and efficient multiplication algorithm, called "duo-multiplication", is proposed. Two multiplications and the sum of their results are calculated directly, with the same number of operation steps as is required for the conventional single multiplication. The algorithm is derived from adjusting non-zero partial product generation timing in order to interleave one multiplication execution into the other multiplication execution. This timing adjustment is realized by utilizing the fact that any two algebraic numbers in two's complement representation can be transformed into a binary signed digit representation pair, in which the non-zero digits of one number do not overlap with those of the other number, without increasing total word length. As duomultiplication operation is fundamental on digital signal processing, the algorithm adoption doubles signal processing capability of a microprocessor.

This paper proposes a video-quality estimation method based on a no-reference model for realtime quality monitoring in video-streaming services. The proposed method analyzes both bitstream information and decoded pixel information to estimate video-quality degradation by transmission errors. Video quality in terms of a mean squared error (MSE) between degraded video frames and error-free video frames is estimated on the basis of the number of impairment macroblocks in which the quality degradation has not been possible to be concealed. Error-concealment effectiveness is evaluated using motion information and luminance discontinuity at the boundaries of impairment regions. Simulation results show a high correlation (correlation coefficients of 0.93) between the actual MSE and the number of macroblocks in which error concealment has not been effective. These results show that the proposed method works well in reatime quality monitoring for video-streaming services.

This paper presents a no-reference (NR) based video-quality estimation method for compressed videos which apply inter-frame prediction. The proposed method does not need bitstream information. Only pixel information of decoded videos is used for the video-quality estimation. An activity value which indicates a variance of luminance values is calculated for every given-size pixel block. The activity difference between an intra-coded frame and its adjacent frame is calculated and is employed for the video-quality estimation. In addition, a blockiness level and a blur level are also estimated at every frame by analyzing pixel information only. The estimated blockiness level and blur level are also taken into account to improve quality-estimation accuracy in the proposed method. Experimental results show that the proposed method achieves accurate video-quality estimation without the original video which does not include any artifacts by the video compression. The correlation coefficient between subjective video quality and estimated quality is 0.925. The proposed method is suitable for automatic video-quality checks when service providers cannot access the original videos.

This paper describes low-power architecture-methodologies for programmable multimedia processors, which will become major functional units in System-On-a-Chip. After brief review on multimedia processing and low-power considerations, recent programmable chips, including MPUs and DSPs, are investigated in terms of low-power implementation. In order to show the difference of the low-power approaches between programmable processors and ASIC processors, a single-chip MPEG-2 encoder is also included as an example of ASIC design.

This paper proposes a video-quality estimation method based on a reduced-reference model for realtime quality monitoring in video streaming services. The proposed method chooses representative-luminance values for individual original-video frames at a server side and transmits those values, along with the pixel-position information of the representative-luminance values in each frame. On the basis of this information, peak signal-to-noise ratio (PSNR) values at client sides can be estimated. This enables realtime monitoring of video-quality degradation by transmission errors. Experimental results show that accurate PSNR estimation can be achieved with additional information at a low bit rate. For SDTV video sequences which are encoded at 1 to 5 Mbps, accurate PSNR estimation (correlation coefficient of 0.92 to 0.95) is achieved with small amount of additional information of 10 to 50 kbps. This enables accurate realtime quality monitoring in video streaming services without average video-quality degradation.

This paper provides a performance evaluation of our audio and video CODEC by using a method for rapidly verifying and evaluating overall performance on real-time workloads of system LSIs integrated with SPXK5SC DSP cores. The SPXK5SC have been developed as a DSP core well-suited to system LSIs. Despite the fact that it is very important to evaluate the overall performance of target LSIs on real workloads before actual LSI fabrication, software simulators are too slow to deal with real workloads and full hardware prototyping is unable to respond well to design improvements. Therefore, we have developed a hardware emulation approach to be used on system LSIs integrated with a SPXK5SC DSP core in order to evaluate the overall performance of audio/video CODEC on a target system. Our emulation system using a DSP core TEG, which has a bus interface, and an FPGA is suitable for overall system evaluation on real-time workloads as well as architectural investigation. In this paper, we discuss the use of the emulation system in evaluating performance during AV CODEC execution. In addition, an architecture design based on our emulation system is also described.

A novel coding scheme using orthonormal wavelet transform is proposed. Various forms of transform coding and subband coding are first reviewed. Then a wavelet coding method is proposed adopting a new approach similar to the one used for transform coding. The approach differs to conventional ones which considers wavelet coding as a class of subband coding. Simulation work is carried out to evaluate the proposed coding method. Significant improvement is obtained in subjective quality, and some improvement is also obtained in signal to noise ratio. Wavelet coding is still in its early stage of development, but can be considered to be a promising technique for image coding.

This paper describes a multiresolutional Gaussian mixture model (GMM) for precise and stable foreground segmentation. A multiple block sizes GMM and a computationally efficient fine-to-coarse strategy, which are carried out in the Walsh transform (WT) domain, are newly introduced to the GMM scheme. By using a set of variable size block-based GMMs, a precise and stable processing is realized. Our fine-to-coarse strategy comes from the WT spectral nature, which drastically reduces the computational steps. In addition, the total computation amount of the proposed approach requires only less than 10% of the original pixel-based GMM approach. Experimental results show that our approach gives stable performance in many conditions, including dark foreground objects against light, global lighting changes, and scenery in heavy snow.