encoding control means for, when said unit group includes a plurality of different types of pictures which are to be encoded with different encoding methods, setting a target quantizer step size used to encode each of the different types of pictures included in said unit group, and for performing a control operation to generate and furnish an updated quantizer step size for each of said I, B and P pictures so that a ratio among the target quantizer step sizes set for the different picture types is a predetermined one, said control operation not being totally depending on the allocation of quantity of the target amount of codes based on the global complexity measure for each of the picture, but in accordance with features of the sequence of moving pictures which represent a degree of complexity of the sequence of moving pictures to be encoded,

wherein if the extracted feature of said sequence of moving pictures indicates that the amount of motion between pictures is relatively small, said encoding control means sets the amounts of generated codes assigned to each I-picture, each P-picture, and each B-picture within said unit group so that the amount of generated codes assigned to each I-picture is the largest, the amount of generated codes assigned to each P-picture is the second-largest, and the amount of generated codes assigned to each B-picture is the smallest, and, as the amount of motion between pictures represented by the extracted feature increases, updates said ratio among the target quantizer step sizes for the different types of pictures so that the differences among the amount of generated codes assigned to each I-picture, each P-picture, and each B-picture are reduced; and

encoding means for encoding said each picture included in said sequence of moving pictures including said each picture using said quantizer step size furnished by said encoding control means and using either said each picture or prediction from a past intra-coded image and/or a predictive coded picture,

wherein said encoding control means initially sets the quantizer step size for a macroblock to be encoded first in said each picture currently being encoded to the target quantizer step size set for the picture type of said each picture currently being encoded, and then, each time said encoding control means encodes each of macroblocks remaining in said each picture currently being encoded, updates the quantizer step size initially set for the first macroblock so that the average of the quantizer step sizes used during the encoding of all macroblocks in said each picture finally approaches the target quantizer step size set for the picture type of said each picture currently being encoded.