摘要 |
PROBLEM TO BE SOLVED: To reduce an overhead due to a motion vector or the like without deteriorating image quality. SOLUTION: Prediction errors Eg1 (i=1, 2,...N) are calculated by plural mounted prediction methods (step 11). A minimum prediction error Egmin is calculated from the prediction error Eg1 (step 12). A quantization step Q of a coding object small block is calculated (step 13). A permissible range T is calculated from the quantization step Q (step 14). A prediction method that satisfies a relation of Eg1 <=Egmin +T is detected by the prediction errors Eg1 of each prediction method (step 15). A prediction method in which a overhead is smallest among the detected prediction methods is decided as the prediction method for a coding object small block. |