Yuzo Senda

A simplified motion estimation using an approximation for the MPEG-2 real-time encoder

We propose a simplified motion estimation method which provides motion vectors for all types of motion compensation used in MPEG-2. The method is a result of applying a newly introduced approximation to the canonical three-step method described in the MPEG-2 Test Model. It reduces the number of necessary computations in the second and the third steps to less than 1%, and that of data transfers to about 8% of the canonical method. Total complexity of the proposed method is nearly that of the full-pel search motion estimation.

Detented Quantization to Suppress Flicker Artifacts in Periodically Inserted Intra-Coded Pictures in H.264 Video Coding

This paper describes a new flicker suppression method for periodically inserted intra-coded pictures (I-pictures) in H.264 video coding. An H.264 encoder periodically inserts I-pictures for channel hopping and random access. The coding noise pattern for I-pictures differs from that of previously appearing Predictive-coded Pictures (P-pictures) because, unlike with P-pictures, inter-frame prediction is not used with I-pictures. This discontinuity in coding noise patterns generates intra-flicker and heavily degrades subjective video quality at low bit rates. We propose detented quantization (DQ) to reduce the discontinuity in coding noise patterns between P-pictures and I-pictures. DQ stabilizes the representation levels of input coefficients in I-pictures on the basis of a derivation of those of the inter-coded images produced from previous P-pictures. Simulation results show that DQ reduces intra-flicker by more than 50% in H.264 video coder JM8.6, and it significantly improves subjective video quality.

A realtime software MPEG transcoder using a novel motion vector reuse and a SIMD optimization techniques

A realtime software MPEG transcoder has been developed. Novel motion vector reuse and SIMD optimization techniques are introduced to accelerate the transcoder without any quality degradation. Mean absolute error approximation criteria are employed in the reuse technique to refine scaled motion vectors. The developed transcoder on a Pentium II 266 MHz machine runs 2.5 times as fast as realtime, when scaling an MPEG-1 bitstream to half size.

Acoustic event detection based on non-negative matrix factorization with mixtures of local dictionaries and activation aggregation

This paper proposes a new non-negative matrix factorization (NMF) based acoustic event detection (AED) method with mixtures of local dictionaries (MLD) and activation aggregation. One of the key problems of conventional NMF-based methods is instability of activations due to redundancy of a region spanned by the bases of dictionaries. Sounds inside the redundant region are often decomposed into undesired combinations of bases and activations that cause failure of detection. The proposed method employs MLD for allocating sub-groups of basis dictionaries to acoustic elements to minimize redundancy in the region and obtain controlled activations. In order to make activations more stable, the proposed method also introduces activation aggregation which combines basis-wise activations into acoustic-element-wise activations. Much more stable activations by the proposed method lead to significant improvement in F-measure by up to 60% compared to an ordinary convolutive-NMF-based method. The proposed method also outperforms a latest alternative which is not based on NMF.

Theoretical background and improvement of a simplified half-pel motion estimation

This paper describes theoretical background of a mean absolute error (MAE) approximation method which was adopted as a half-pel motion estimation method in our real-time MPEG-2 codec VisuaLink 7000. We also propose an improved MAE approximation method which employs the horizontal and the vertical differentials of a source picture in addition to the MAEs of adjacent full-pel motion vectors. In general, any approximation method for MAE involves some loss of coding efficiency. However, the proposed method reduces loss of coding efficiency to 1/2/spl sim/3/4 of other existing methods with negligible increases of necessary computations and data transfers.

Accurate video-quality estimation without video decoding

This paper proposes a video-quality estimation method which does not need video decoding. It takes into account error-concealment effectiveness and the location of video-quality degradation to estimate video-quality from only packet-header information. Error-concealment effectiveness is evaluated on the basis of motion-level information estimated by the data-size of each frame. The quality-degraded location in a frame is evaluated by the transmission-error timing. Experimental results show that a high correlation coefficient of 0.853 with a lower outlier ratio is achieved with the proposed method. The method is suitable for a low-complexity video-quality monitoring system in IPTV services due to its accurate video-quality estimation without video decoding.

Fast Compression Implementation for Hyperspectral Sensor

Fast and small foot print lossless image compressors aiming at hyper-spectral sensor for the earth observation satellite have been developed. Since more than one hundred channels are required for hyper-spectral sensors on optical observation satellites, fast compression algorithm with small foot print implementation is essential for reducing encoder size and weight resulting in realizing light-weight and small-size sensor system. The image compression method should have low complexity in order to reduce size and weight of the sensor signal processing unit, power consumption and fabrication cost. Coding efficiency and compression speed enables enlargement of the capacity of signal compression channels, which resulted in reducing signal compression channels onboard by multiplexing sensor signal channels into reduced number of compression channels. The employed method is based on FELICS1, which is hierarchical predictive coding method with resolution scaling. To improve FELICSs performance of image decorrelation and entropy coding, we applied two-dimensional interpolation prediction and adaptive Golomb-Rice coding, which enables small footprint. It supports progressive decompression using resolution scaling, whilst still delivering superior performance as measured by speed and complexity. The small footprint circuitry is embedded into the hyper-spectral sensor data formatter. In consequence, lossless compression function has been added without additional size and weight.

Fast software MPEG-2 video transcoder with optimization of requantization error compensation

A low-complexity software video transcoder with motion compensation (MC) of requantization errors has been developed. For reducing the processing time of the developed transcoder, two novel methods are proposed. One is a sparse block transcoding method that bypasses a transcoding process according to the distribution of the transform coefficients in an input block. This method is valid for about half of all blocks of P pictures. The other is a fast MC method that alternates rounding directions of MC per picture to decrease conditional judgments. The processing time of the fast MC is reduced to three fifths. The transcoder with the proposed methods has been running in real-time on PCs.

In-loop noise shaping based on pseudo noise injection and Wiener filtering

This paper proposes an in-loop noise shaping method based on the combination of pseudo noise injection and Wiener filtering for High Efficiency Video Coding (HEVC). In deblocking process, the proposed method injects pseudo noise into the vicinities of deblocked edges where signal-dependent coding noises are supposed to appear. The pseudo noise injection introduces a masking effect of signal-dependent noise with signal-independent noise. Since subsequently applied Wiener filtering accomplishes optimal noise reduction in a minimum-mean-squared-errors sense, it minimizes the deleterious impact of the pseudo noise injection on coding performance. Simulation results using HEVC Test Model software show that the proposed method successfully suppresses banding artifacts with a negligible impact on PSNR values, bit rates, and encoder/decoder runtime measures.

Enhanced reconstruction of AVC/H.264 intra video based on motion compensated temporal filtering and total-variation regularization

This paper proposes an enhanced reconstruction method of AVC/H.264 intra video based on Motion Compensated Temporal Filtering (MCTF) and Total-Variation (TV) regularization. The MCTF attenuates temporal fluctuations in block transform coefficient values of individual decoded video frames on the basis of Motion Estimation (ME). However, the attenuated block transform coefficients lead to spatial discontinuities on the associated image area. The TV regularization technique reduces the spatial discontinuities, and also suppresses quantization noises of the image edge vicinities that are supposed to belong to a smooth object while keeping the image edge sharpness. The proposed reconstruction method thus enhances the quality of decoded video frames degraded by flicker artifacts and ringing noises. Simulation results show that the proposed method improves the subjective video quality of AVC/H.264 intra video, and the attainable bit rate up to 16.1% at the same objective video quality.