| 摘要 |
<p>A digital halftoning mechanism (100) has blue noise characteristics to significantly improved perceived image quality while significantly decreasing computational requirements. Halftoned binary images are produced from grayscale images by taking the input image a region (150) of pixels at a time, and computing thresholds for every pixel in that region (150), such that the locations of thresholds close in value are as random as possible consistent with being anticorrelated. The method can be applied on-the-fly so that a threshold is located, used and discarded with no need to store it in a mask. Because there is a fresh start for every region (150), it is extremely unlikely that any periodicity due to the halftoning will appear in the output. An alternative way to perform the halftoning process is to compute an anticorrelated thresholding mask and then apply the mask to the image by regions (150). In another alternative, instead of a mask, a list of the locations of the thresholds in the region (150) is stored, addressed by threshold value, and applied to the region (150) of the image by going through the thresholds sequentially. Multiple realizations of the mask or list may be computed, stored and chosen randomly. Quadtrees may be used to address a square region. Alternatively, a "parquet tile" method of recursively addressing a square region may be employed. <IMAGE></p> |
