A method of quantizing pixel values in an image formed by a plurality of pixels, each pixel representing an optical density of the image at a location within the image, and having an original optical density value selected from one of a set of `c` original optical density values that has a number of members larger than a desired output set of `d` optical density values through a process of combined halftoning and cell-to-cell error diffusion. Initially, each pixel is modified with a pro rata portion of an error term determined for a previous halftone cell, or from the current cell in a previous interaction. A set of modified pixels, corresponding in position to an m.times.n halftone cell, are processed in accordance with a standard halftoning process, to derive a set of output pixels at one of d optical density levels. The calculated output density is used to address a lookup table having a stored set of empirical density values upon printing the dot represented by a halftone cell with the same count of optical density levels. The empirical density value for the cell returned from the lookup table is compared to a sum of the modified pixel values for the cell. The difference (which represents the difference in gray level input and the actual output) is directed on a weighted basis to the pixels of subsequently processed dots.