Method of and apparatus for compressing image data

Image data (fij) depicting an image in a pixel block are converted by 2-D Discrete Cosine Transformation into transformation coefficients (Fmn or Fmn*). Only a part of the coefficients are employed to form compressed image data (Dt). The selection of the part of the coefficients is executed using code tables (CT). Several groups of code tables are prepared, and one of those groups is selected in the data compression operation in response to a standard deviation (σ) of the image data in the pixel block. When the compressed image data is restored, identification data specifying the selected group in the previous data compression is stored in the restored image data. Then, in the next data compression of the restored image data, the selected group is again selected as a function of the identification data.