A method is disclosed for controlling the amount of compressed data. One or more viewing condition parameters are specified, and a quantizer step size for each frequency subband is determined that minimizes an amount of data that must be encoded for a compressed representation of the image by using one or more viewing condition parameters and a model that characterizes the human visual system. The image is compressed using the quantizer step size for each frequency subband to produce an amount of compressed data. The presence of a pre-specified maximum amount of compression data is determined. The amount of compressed data is compared to the pre-specified maximum amount. Subsequent discarding of least visually relevant compressed data reduces the amount of compressed data.