The weighted sum of the absolute values of the transform coefficients, defined herein as the activity index, is proposed as an objective measure of scene busyness (i.e., the density of significant scene detail). For an image divided into subpictures, it is possible to classify each subpicture into a finite number (say four) of categories according to its computed activity index. A different coding scheme, involving different truncation and quantization rules and hence a different number of bits, is used for each activity category. Data compression is efficiently achieved by assigning more bits to code those portions of the image showing the most detail.