Screen-Aware Compression of Binary Image Data
Publication Date: 2005-Nov-08
The IP.com Prior Art Database
When a continuous tone image is to be printed it must usually be "screened" to produce a halftone image consisting of "on" or "off" (binary) pixels. The goals of a good screen generally are contrary to the best conditions for compression of digital data. This disclosure describes a compression method requiring relatively few resources and providing superior compression ratios for screened images relative to other compression approaches.
When a continuous tone image is to be printed it must usually be “screened” to produce a halftone image consisting of on or off (binary) pixels. The goals of a good screen generally are contrary to the best conditions for compression of digital data. The screen essentially imposes “noise” on the image data. Traditional data compression techniques such as Lempel-Ziv-Welch orLZW usually cannot achieve good compression ratios with such data. Typical solutions provide varying degrees of compression with various performance or hardware costs, including the possibility of using a larger memory array and avoiding compression entirely.
In many cases the screen produces a specific repetitive pattern in the binary image and this pattern consists of fairly simple groups of zero bits and one bits. This invention defines a “standard format” for groups of pixels. The size of the group is defined by the horizontal pitch of the halftone screen. Adjacent groups are then compared. In most cases the difference between them will be very small and can be expressed by a code consisting of a few bits. Adjacency can be either horizontal or vertical if the screen allows it. Vertical adjacency can produce better compression, but requires memory to store the previous raster line for reference in compressing or decompressing the current raster line. The technique would generally be augmented with simple run-length encoding to take advantage of the greater compression possible with this technique in regions of the image that are entirely white or entirely black.
The essential concepts of the invention are:
- To examine the source data in bit groups that are the same size as the screen pitch in the scan axis (Note: the technique applies only to images that have been screened with a suitable pattern.)
- To recognize and encode “standard format” groups of pixels containing exactly 1 or 2 transitions
- To effectively separate (and compress with different techniques) the halftoned and non-halftoned regions of the image
Using the screen pitch as pixel group size results in a natural pattern in halftoned regions. Using the same group size for the entire image simplifies the process, although compression might be improved in non-halftoned regions by using a different group size (including, possibly, single pixels). The groups need not be explicitly aligned with the halftone screen because the standard format applies regardless of alignment for the defined set of screens that may be effectively used with this technique. Therefore, it is not necessary to adjust the alignment at the beginning of each raster line. That is, the examination of successive groups can proceed across raster line boundaries without extra steps.
Non-halftoned regions are loosely distinguished by the presence of relatively long runs of “0” bits or “1” bits and these would typically be compressed with run-length e...