My current image editing software is Aldus (now Adobe) PhotoStyler (for Windows 3.1 and no longer made - I am a sort of Luddite, and not overly proud of that). I use that for making the stereo pairs, too.
To combine images and make them into anaglyphs, I've started using Gunter Richter's "Images 3D", version 1.6. This program has a rotation control feature that I've found to be essential.
Then comes the processing, and here's where you can do a lot to help or damage a picture...
 |  | The picture frames have to be oriented correctly. It's often very difficult or impossible (especially on hyperstereo pictures) to have both pictures level relative to each other. So... get that right, and if you can determine it, rotate the frame that is least parallel to the dolly plane, too (if you can't determine which, that's a good sign). |
 | Make sure the two images are at the same scale. Reduce the larger one - do not enlarge the smaller one! Software algorithms are much more likely to produce good results if allowed to lose information, than interpolating new information. However, both rotating and resizing operations will cause a loss of information (fuzzing or blurring). If rotation of a frame is necessary, consider overriding the decision regarding which frame to resize, choosing to both rotate and resize the oversized picture instead. It would be best if you got both shots aligned and sized right at the time you take the pictures. The better you are at that, the less you have to dink around in here. I do the resizing in PhotoStyler. | |
 | Edit any differences of movement or objects not in both frames. Perhaps there's a bird in the sky. Either add the bird (difficult, because you have to figure where to put it, and I'd call up the second frame and try to align it while parallel viewing both pictures to try that), or edit out the bird. For scans of paper photo pairs, watch out for dust specks and film anomalies, etc. too. They may look small here, but they'll glare at you in the final anaglyph. Don't go crazy with this, like trying to edit traffic. Cars are a nuisance in real life, why not in photos, too? Pick your battles. | |
| Save the edited frame(s). Save in a non-lossy format, such as .BMP or .TIFF. Do not save as a .JPG. There's more work to be done, and you do not want to corrupt the image in any way (yet). | ||
 | Align the frames vertically. Track corresponding pair points in the picture where red and cyan parts stand out, such as a rock, or rooftop, or plant part. Align to correct for vertical displacement. If you did your resizing and aligning tasks correctly, you should see no (or just one to three or so pixels) change in vertical displacement from points at the top of the picture and points at the bottom. | |
 | Align the frames horizontally for correct spacing and depth.The official rule is, no part of the picture should come forward of the frame. Theory is great, but I find that red/blue glasses come in a wide variety of shades of blue and/or cyan and/or green. That means the colors don't properly filter, and ghosts can become a problem. That's when I punt, and go for making the object of interest the convergence point for the red and cyan (which places it at frame depth). | |
| This is when you put your 3D glasses on and test-fly the picture. Use the horizontal adjustments until the closest point is at frame depth (the fourth frame in the .GIF movie above), and the rest further away. Do you see ghosts next to the object of interest? If so, you might consider breaking the official rule. I would prefer to see the object of interest be in good focus/alignment than worry about whether my picture satisfied some unknown person's criteria. However, if you want to have your pictures taken seriously by the stereo purists, you will need to consider this point.
Put your glasses on and watch the above animation. You'll notice that at the fourth frame, the smiley suddenly becomes very sharp. If you place your mouse next to the face, you'll also be able to see how the face's depth changes with the alignment. | ||
| Save the image. Mostly, I like black-and-white versions ("Gray Red-Cyan" in Anaglyph Maker), because color versions usually end up with my eyes conflicting over which frame gets priority (because of the color filtering). But, if a color anaglyph version ("Color Red-Cyan") looks good, I'll save it as well (not instead of). That's just me. | ||
 Adjust gamma, or brightness/contrast. This is done in your image editing software (PhotoStyler's "Gray/Color Correction"). Commonly, the brights are dimmed, so I slide up the gamma a little.
| ||
| Resize? You may want to resize the image at this point - downward, if at all. Do not enlarge - the results simply won't be as good, unless you're running some super-spiffy software that I've never seen. I run at 1024x768, so I often resize images to fit my screen. Of course, lots of viewing softwares allow you to resize on the fly, but I find that they leave chunky artifacts from substandard algorithms. Image editing software seems to do a much better job of resizing. Save both versions. Hard disk space is cheap, relative to the efforts you're putting into making nice pictures! | ||
| Sharpen areas of intricacy. I select areas (using a "magic wand" selection tool) that are not sky, snow, or other areas of very low contrast, and then apply a "sharpen" function to them. With the glasses on, I hit Alt-Backspace ("undo" and "redo" toggle in PhotoStyler) while inspecting all parts of the picture. If the sharpening seemed to help, I keep it. If not, I leave it undone. Do not sharpen until after you've resized the image (if you intend to resize at all). | ||
 | Crop the image? Quite often there'll be white areas left over from the anaglyph alignment process. Crop that crap. Aside from the 1024x768, I like to shoot for cropping images to the nearest 10 pixels, if the picture ends up still looking good. It's just a neatness thing - it isn't important. I will not resize an image to get round numbers, unless it's part of resizing for other reasons, too. | |
| Save the image - you're done editing. I still save in an uncompressed format. I don't delete this file until I'm happy that the .JPG (next step) looks good. | ||
Convert the image to .JPG. I use ACDSee to convert and save into .JPG format. Also, some settings may need to be changed. Ed K gave me the parameters, and I ended up with these:
| ||
The darks are usually too dark, but you also need to maintain black, so I edit using PhotoStyler's "Grey/Color Map" option, which allows me to select pieces of the brightness scale. I move the dark end up, and let it drop quickly to black.