Method and apparatus for adaptive transform coding of picture signals

• US 4,225,885 A
• Filed: 02/13/1978
• Issued: 09/30/1980
• Est. Priority Date: 02/12/1977
• Status: Expired due to Term

• Current Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Original Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Inventors: Lux, Peter, Petzold, J/rgen
• Primary Examiner(s): Boudreau, Leo H.
• Application Number:
  US05/877,382
• Time in Patent Office: 960 Days
• Field of Search: 340/146.3 MA, 340/146.3 AG, 340/146.3 H, 340/347 AD, 356/71, 358/260
• US Class Current: 382/239
• CPC Class Codes: H04N 19/154 Measured or subjectively estimated visual quality after decoding, e.g. measurement of distortion use of rate-distortion criteria H04N19/147
• CPC Class Codes: H04N 19/60 using transform coding

Abstract

A method and apparatus to reduce the redundancy for the transmission or storage of picture signals using transforms. The basis coefficients are arranged in a sequence and thereafter quantized in the customary manner. The basis coefficients obtained from the transformation are sequentially examined to see if they exceed a predetermined value, for example the average value across the entire picture. The number of times the value is exceeded is counted. If, in a sequence class, the number of coefficients exceeding the predetermined value is more than half the total number in the sequence class then the associated sequences are quantized with a larger number of bits. For this purpose, activity classes are formed which indicate which sequences in the picture are active and the characteristic of the quantizer is controlled by this class information.

Claims

1. A method for adaptive transform coding of picture signals comprising the steps of:

transforming the picture signals in accordance with a unitary orthogonal transform to obtain a plurality of coefficient value signals representative of picture energy in a series of subpictures, the coefficient value signals being grouped into a plurality of sequence classes, each sequence being associated with a distinct, fixed threshold signal value;

comparing each of the coefficient value signals with its associated threshold signal value and counting the number of coefficient value signals in each sequence class which exceed the associated threshold signal,

quantizing said coefficient value signals, the number and range of quantization levels applied to each coefficient value signal being determined by the combinations of the number of coefficient value signals in each class which exceed the associated threshold signal value.

2. A method as claimed in claim 1 in which the picture is divided into subpictures and the coefficient value signals are obtained by the consecutive transformation of subpicture signals and wherein the quantization value ranges for each subpicture are independently determined.

3. A method as claimed in claim 1 or 2 wherein the threshold signal values associated with a sequence class represent the standard deviation of the coefficient value signals in that sequence class as measured in a plurality of representative pictures.

4. A method as claimed in claim 1 or 2 wherein the coefficient value signals are normalized prior to said quantizing.

5. Apparatus for adaptive transform coding of picture signals comprising:

unitary orthogonal transform means connected to receive subpicture signals and to produce sequences of coefficient value signals therefrom, the coefficient value signals being grouped into a plurality of sequence classes;

first memory means which store and supply a plurality of threshold signal values, each sequence of coefficient value signals being associated with a threshold signal;

comparator means connected to receive coefficient value signals from the transform means and the associated threshold signal value from the first memory means and which supply a reference signal whenever a coefficient value signal exceeds the associated threshold signal value;

first counter means connected to count the number of reference signals produced from coefficient value signals in each sequence class and to produce an activity signal whenever a predetermined count is exceeded for each sequence class;

second memory means connected to receive and store the activity signals associated with each sequence class;

logic means connected to receive and logically combine the activity signals derived from a subpicture signal, to produce quantizer control signals from the logical combination of signals, and to store the quantizer control signals; and

quantizer means connected to receive the coefficient value signals produced from a subpicture signal together with the associated quantizer control signals and to quantize the coefficient value signals in accordance with a quantization characteristic which is uniquely determined by the quantizer control signals.