ADAPTIVE SELECTION OF QUANTIZATION SCALES FOR VIDEO ENCODING

The quantization scale selected for encoding the current frame of a video sequence is selected based on a metric generated based on a set of image data in the video sequence. For example, in MPEG encoding, the linear quantization scale is selected for use in encoding the current frame if the average quantization level used to encode the previously encoded frame is between specified high and low thresholds. Otherwise, the non-linear quantization scale is selected. As a result, medium-difficulty sequences will tend to be encoded using the linear quantization scale, while low- and high-difficulty sequences will tend to be encoded using the non-linear quantization scale. For most normal video sequences, this will result in fewer incidents of panic mode video compression processing and improved picture quality.