To enable transmission using a smaller number of bits than usual by making it possible to encode an image of the same picture quality as conventional quantization without aggravation in overload noise with a quantization characteristics having a smaller dynamic range when an adaptive quantizer is applied to a quantizer for predictive error orthogonal conversion encoding.

The quantizer 10 of the adaptive quantizer 3 obtains a quantization output Q quantized according to quantization characteristics from the output Et of an orthogonal converter 2 and also levels Q+ and Q- above and below Q. One of them is selected as a temporary quantization output by a selector 11 and reversely converted by a reverse orthogonal converter 5t, and the amplitude value of the converted signal is found to decide that a conversion coefficient having a large amplitude value is affected considerably by quantization noise; and the quantization step is made to smaller one or larger one and a decision controller 17 decides whether or not a temporary local decoded signal exceeds the dynamic range, so that the said decision process is repeated when the dynamic range is exceeded.