rvklassen@watcgl.UUCP (01/16/87)
There have been several requests for a scheme whereby a bit-mapped image can be resampled at a higher resolution for a printer. In addition there were a number of suggestions of methods of changing the resolution of raster images. Unless the original requestor and I read the term Bit-Mapped differently, the responses so far have been to a different question. There are really two problems that one can ask when talking about resolution changes for bit-mapped images. The first one is how to change the resolution of an image that is really black and white (like a bit-mapped font.) For this you have to do edge detection and somehow decide what the shape of the `ideal' edge is before re-drawing it at the new resolution. I assume this is not the question the original poster had in mind. The second (and I imagine pertinent) question, is what to do with an image that looks like grey-scale, even though it is only one bit per pixel. With such an image the `greyness' of a region is given by the number of black pixels in that region. You somehow have to decide how large a region of the source image to consider, and then move a window over the source image, counting the black pixels, and then produce a region in the final image of the same `greyness' using either a dither or half-tone. Foley and Van Dam `Fundamentals of Computer Graphics', chapter 12, can provide an introduction to dithering and half-toning, and has references for further reading. When you count the black pixels in a region, you can count pixels that are away from the centre as worth less than one: the set of weights they have is called the filter kernal. The simplest is just the rectangular filter (all the same weight - just count them), myriad other kernels are discussed in the signal processing literature. There are two independant problems here: finding a good set of grey levels and displaying them on the higher resolution bit-mapped device. Because you are stuck with one bit input and one bit output there are no perfect solutions. If all you have is those two devices then you can't test your solution to one problem without seeing the effects of the solution to the other problem. If you only have one pair of resolutions you can select the weights and half-tone dots by hand. regions differently,