In current H.264/AVC, it is able to use either inter prediction coding or intra prediction coding for one macroblock only. Yet, if a size of a macroblock having a size of 32×32 greater than a current 16×16 or a size over 32×32, it is able to achieve high compression performance using both of the inter prediction coding and the intra prediction coding for one macroblock. By enabling sub-macroblocks within 32×32 macroblock to have sizes differing from one another, i.e., by adopting a plurality of intra prediction modes of sub-macroblocks within 32×32 macroblock, high compression performance can be achieved. Yet, referring to FIG. 23, it is able to achieve higher compression performance by predicting sub-macroblocks within one macroblock in a manner of mixing the inter prediction and the intra prediction together.

Meanwhile, if a macroblock includes a sub-macroblock coded using intra prediction and a sub-macroblock coded using inter prediction, a prediction mode flag indicating whether each of the sub-macroblocks is inter-predicted or intra-predicted is transmitted for each sub-macroblock. Moreover, it is able to transmit inter prediction information (e.g., reference picture index information, motion vector information, etc.) for the inter prediction coded sub-macroblock and prediction direction information for the intra prediction coded sub-macroblock.

Meanwhile, by transmitting DC transform size flag information indicating a DC transform size, as mentioned in the foregoing description with reference to FIG. 19, it is able to specify a prediction mode of the intra prediction coded sub-macroblock.