This report is concerned with the bit-rate reduction of signals produced by Walsh-Hadamard transformation of PCM television signals. The main characteristics of the transform signals are outlined and the methods of linear and non-linear coding used to reduce their bit rate are described. It was found that, with linear coding, a slightly higher bit rate is required to describe the transform coefficients than is needed for the original signal. However, using non-linear coding, the data rate can be reduced to 7.5 bits per sample with no visible impairment. For operation at lower bit rates, it is an advantage to sub-sample the signal at colour subcarrier rate before transformation, as this allows more efficient encoding of colour signals. Then, at 5.5 bits per sample, although many pictures are reproduced without impairment, those containing high-contrast edges or saturated colours are noticeably impaired. The results of comparative subjective tests show that the bit-rate reduction performance of the transform system is not as good as that of differential PCM. However, if the effects of channel errors and the additions to the digital signal necessary to protect against them are considered, the advantage of DPCM is partially offset.
