Human Systems Integration Division @ NASA Ames

One way to describe the problem of digital half toning is as a search for the quantized image that minimizes the visibility of artifacts. To apply this approach in practice, it is first to specify a computational model for computing visible error that can be used to rank candidate images automatically. The model may be incorporated directly into a search algorithm, or used after-the-fact to rank images produced by algorithms that are more heuristic. These approaches have been quite successful when applied to achromatic images, even when using a relatively simple visual model accounting for high-frequency contrast sensitivity but with no masking. This approach can be directly generalized to additional dimensions such as time and color. Visual models based on threshold measures may not be optimal for low bit rate conditions where quantization noise is visible. Instead, the degree to which the noise is effortlessly segmented through perceptual scission may influence the utility of the final image.