Abstract: Variable length-adaptive image data compression method and apparatus. The statistical characteristic of a frame containing a number of blocks is extracted by blocks and then encoded, so as to achieve a bit distribution for each block. An optimum quantizer is selected which can compress data to a quantity most approximating to the quantity of bits allocated to the AC coefficient of each block. The present invention adopts a variable length coding for providing a high picture quality, but limits the number of classes containing the same number of blocks to a number (for example, 4) predetermined based on the statistical characteristic of input image data. Accordingly, it is possible to easily control the quantity of data and to converge errors into blocks.