In interframe coding the content of the frame memory can also be utilized for the prediction of the luminances of moving area picture elements (pels), provided that their displacement has been measured and transmitted. After an analysis of methods for the optimal segmentation of the image into areas corresponding to background, translating objects, and areas not predictable from the previous frame, experimental results of the coding are given; it can be seen that the bit rates are about halved with respect to plain interframe coding.

In frequency demodulation threshold noise appears in the form of pulses of various amplitude and width. When a TV signal is transmitted these pulses can be detected at the outputs of a line-by-line subtractor, since their amplitude is in general much greater than the amplitude of the noise and of the residual video signal. Based on this principle a threshold noise suppressor (TNS) has been applied at the output of a phase-lock loop (PLL) demodulating a video signal. The TNS consists of two 64-μs delay lines and a simple logic that detects the spikes and substitutes for the signal the average between the previous and following line when a spike is present. The major and practically unique limit to the operation of the TNS is due to spikes interference, i.e., the simultaneous appearance of two pulses at the input of either line-by-line subtractor. The comparison could be made in principle between more lines with better results. However with the device consisting of only two delay lines the original threshold is lowered at least 3 dB. The same principle is also applicable to a color video signal. In the paper, details of the theory, the implementation, and the limits of the TNS are given together with experimental results.