CompuServe Messages

You can use any Amiga image processing program, such as the $70 Imagemaster R/t from Black Belt Software, to either alter each image to conform to the first, or to do a soft-edged composition when stitching the 4 images back into one. I suppose that the differences come from the scanner's automatic brightness and contrast adjustments, i.e. the red, green and blue components are affected equally. (If not, get a new scanner — this one would alter hues if it applied different operations to R, G and B) Here's how I would do color correction using Imagemaster: say image 1 is the top left scan, and image 2 is the top right. I would use "Define color" to get RGB values from a few distinguishing marks on the right side of image 1, and get RGB values for the same spots on image 2. Using green values for example, I would plot value pairs on graph paper, with readings from image 2 lying along the vertical (Y) axis, and those from image 1 along the horizontal (X) axis. After plotting a few points, you should see a trend emerge, hopefully in the form of a straight line. If so, then you can apply one contrast and one brightness correction to image 2, and make it match image 1. Measure the slope of the line; this tells you how much to increase the contrast in image 2 to match image 1. When the contrast operation is complete, take some new measures and draw a new graph; the line that appears should have a slope of 1, but it may not intersect the Y axis at the origin (0,0) point. If so, you need to add or subtract a constant value to all the pixels in image 2, and they'll finally match those in image 1. Just look at the Y coordinate where the line touches the Y axis, and subtract that amount from image 2 Whew! Here's an example. I clicked on the same 3 spots in images 1 and 2. The green component of each measurement was: 1 2 — — 124 112 233 246 36 4 The slope of the line can be computed by using any 2 points and dividing (difference in "1" values) by (difference in "2" values). In this example, the slope is (233-36)/(246-4) = 1.21. This means we must increase image 2's contrast by .21, or 21%. After doing so, we take 3 more measures: 1 2 — — 124 112 17 3 255 247 If the new line had a slope of 1, there would be a one-to-one relationship between values in image 1 and image 2; the fact this is not true here can be attributed to small errors in calculations, or that you didn't exactly pick the same points in each image. At any rate, the difference between images is around 11 luminance units (out of 256). Imagemaster expects a fractional number, here (11/256) = .043. In Process/Scientific/Add or Subtract a Constant, select Monochrome to add the same value to all three R, G, and B components, and enter .043 as the value. Presto, image 2 matches image 1 as closely as possible. The process may sound really scientific, but it's quite easy to follow if you try to mimic my example. The best part of it is, you never have to guess at how much correction to apply, only to Undo and Redo with different settings. One contrast, one add/subtract, and you're done. Certainly worth a bit of math, if you ask me. Blaq!