To facilitate special reproduction while suppressing increase in a data amount without causing deterioration in the image by converting prediction coding data, which include an orthogonal transformation coefficient and a motion vector resulting from motion compensation of an object image with respect to a reference image and from orthogonal transformation, into in-frame coding data and recording the converted data.

Image data of each frame are processed by a variable length coding decoding circuit 21 and an inverse quantization processing circuit 22 or the like and the result is fed to an adder 24. On the other hand, data decoded by the variable length coding decoding circuit 21 are added to the processed data as above at the adder 24 via a motion vector extract circuit 25 and a motion compensation 26 and output data of the adder 24 are stored in a frame memory 27. The output of the frame memory 27 is fed to the motion compensation circuit 26. Furthermore, the output data from the adder 24 are coded again by a DCT circuit 28 and a quantization circuit 29 and the resulting data are fed to a data recording and reproducing device 37 via a variable length coding circuit 30 and an output buffer circuit 31.