PREDICTIVE ENCODER WITH A NON-LINEAR QUANTIZING CHARACTERISTIC

摘要

Disclosed is a predictive encoder in which undesired redundancy components are effectively suppressed without causing an increase in quantizing noise and a deterioration in picture quality, in the case of video signals. A predictive encoder according to this invention comprises circuitry for producing a predictive signal, a device responsive to an input signal and the predictive signal for producing a predictive error signal representing a difference therebetween, and a quantizer responsive to the predictive error signal and having a non-linear quantization characteristic for producing information codes. The non-linear characteristic for the quantizer is that a transmission gain is set at a value smaller than unity within a sufficiently small amplitude range and at a value substantially equal to unity for amplitude ranges beyond said sufficiently small amplitude range. With this invention, the amplitude distribution for the predictive error signal is concentrated in a smaller amplitude region by passing the predictive error signal through the non-linear quantizer, whereby the information codes can be reduced Even if minute waveform distortion is introduced by the non-linear characteristic, it is hardly detected as deterioration. Further, the quantization noise does not increase, because the quantization steps are made finer than or at least equal to the conventional system. In the composite predictive encoder according to this invention, because the interframe predictive error signal for the stillmotion picture is small, the amplitude thereof is suppressed. Therefore, the noise components are also suppressed. On the other hand, the interframe error signal for a rapidly moving picture, which is of relatively high level, is encoded without resort to amplitude suppression, whereby blur can be prevented.