| 摘要 |
PURPOSE: To reduce the capacity of a memory for storing class prediction coefficients at the time of generating a prediction value by classification adaptive prediction in hierarchical encoding. CONSTITUTION: From the picture signals of a first hierarchy, by thinning filters 2, 3 and 4, the picture signals of second, third and fourth hierarchies with the number of picture elements respectively reduced to 1/4, 1/16 and 1/64 are formed. The picture signals of the fourth hierarchy and difference signals from subtractors 5, 6 and 7 are transmitted. Prediction signals are formed in predicting device 41, 42 and 43 and the predicting devices are provided with a table acquired by learning beforehand. In the table, the prediction coefficients, the prediction values or normalization prediction values are stipulated for respective classes. At the time of deciding the class, the prediction value is formed relating to a picture element under consideration by the linear combination of the class prediction coefficient and the value of a reference picture element. A minimum value of the error between the prediction values calculated for all the classes and the true value of a prescribed picture element is detected and the class corresponding to the minimum value is decided as the class of the picture element under consideration. |
