Studies have been conducted for the development of an HDTV coding and transmission system for contribution purposes. This paper describes the results of investigations into loop filters used in the coding systems. HDTV signals contain a large amount of noise due to their wide bandwidth. This noise leads to poor coding efficiency. A pre-processing of coding input signals, however, is not preferable for high-fidelity transmission. Hence, the introduction of a loop-filter is examined, which reduces the noise in the prediction loop and increases the prediction accuracy. An investigation into a suitable filter characteristics is also conducted. The results show that the loop-filter is very effective for contribution-purpose coding systems, especially for ones using one pel accuracy motion compensation.

View full abstract

The MUSE signal is usually obtained by sub-Nyquist sampling and decoded by re-sampling. Its three-dimensional spectrum, however, is similar to that of the EDTV that the author proposed. MUSE signals can, therefore, be processed by modulating/demodulating three-dimensional sub-carriers; this makes it possible to unify their modems and implement them with various methods, such as the authorized or a simplified method. In this sense, both EDTV and MUSE can be considered to be based on MUlti-dimensional Sub-carriers Encoding("MUSE").

View full abstract

We have examined the signal processing method for MUSE decoder. We applied a 2 dimensional skew-symmetric filter for the video signal processing. This is effective for both reduction in the hardware quantity of the interpolation filter for the moving area, and improvement of the reproduced image quality for the stationary area. Further, we improved the ability of equalization by enlarging the filter length and word length. We have developed a compact MUSE decoder by using the signal processing techniques to make LSIs. As the result, satisfactory performance was obtained as reported herein.

View full abstract

We have developed a new analog bandwidth compression system which can transmit two conventional TV signals through a single analog channel. This system devides the two-dimentional spatial frequency component of television signal into several sub-bands, cuts the high diagonal component, and transmits each sub-band by time division multiplex.

View full abstract

A system for video coding and transmission terminals using a rate conversion technique is presented. The system is capable of transmitting video signals with home-TV quality through low bit-rate (56/64 kbit/s) lines such as the BRI of ISDN and INMARSAT-A HSD services by first storing signals coded at a higher Details are described bit-rate, and then transmitting at the line bit-rate. regarding the rate conversion scheme, coding algorithm, variable rate coding control scheme to improve coding efficiency, pre- and post-filtering techniques to improve coded picture quality, audio and video multiplexing method, and transmission protocols suitable for the various networks. Finally, an outline of two different experimental terminals is given.

View full abstract

This report presents a new motion compensation technique for scaling decoder. The decoder applies inverse DCT of different sizes than that of the encoding DCT and scaling motion vectors accordingly, to reconstruct images that have different dimensions than those of the originally coded picture. The scheme based on the problem that the downsampled motion compensation of existing decoders essentially effects a spatial low-pass filter. Then, we improved the spatial frequency response of motion compensation by modifying the half- and quarter-pel filters. Reconstructed image quality is improved, and as a result, propagation of error is also reduced.

View full abstract

This paper proposes an interframe coding system which uses leaky prediction to eliminate mismatch noise that arises when codecs using different DCT schemes are connected. The system acheives FDCT-IDCT mismatch noise suppression without the insertion of intraframe coded data as proposed in the international coding standard H.261 but complies with all other specsfications. The use of leaky prediction results in loss of coding efficiency and deterioration of picture quality in the image's static areas thus the paper then discusses these effects and proposes solutions which are tested in simulation experiments. We then provide comparative data from these expriments and discuss the codec structure required for this scheme.

View full abstract

In the subband coding scheme, each subband has been encoded independently on the assumption that there is no correlation between subbands. But there is some correlation between subbands caused by following reasons : First, non-stationary of the image has the effect both on the low subband and on the high subband at the same point. Second, the subband analysis filter cannot have ideal cut-off property. So, the pass band of low pass filter and high pass filter has an overlap. We clarified that there is correlation in the local variance between the difference signal of the low subband (or the high passed low subband) and the high subband. Especially, in the case of SSKF, we improved the coding performance 2〜3[dB] in SNR by inter-subband prediction using the difference signal of the low subband

View full abstract

When a color image is divided into small blocks, three color components in each block have a strong correlation with one another. Using Principle Component Analysis, three new color coordinates can be built into each block of the RGB color image. This ensures that data in each new coordinate has no correlation with those in the other coordinates. In general pictures, the variance of a component is much larger than those of the other components. Even with the elimination of the other components picture quality is not degraded.

View full abstract