A method and system for predicting a moving picture to be implemented in an encoder is provided. The system includes a plurality of reference picture memory areas for storing reference pictures to be used for prediction and a prediction generation section that includes a motion compensator and memory updater. The motion compensator receives a parameter representing the motion between an image to be predicted and a reference picture stored in the reference picture memory area and a parameter representing the reference picture memory area to be used for prediction, and generates a predicted image by using the reference picture stored in the reference picture memory area indicated by the parameter representing the reference picture memory area. The memory updater carries out dynamic allocation of reference picture memory areas required for prediction and judges whether to store new reference pictures additionally.