| Author | Message | ||
     Puppet Kite Kid Rating: N/A |
Let me just add what I have done to my anaglyphs, which is different from other techniques... First of all, please play them _only_ on a computer, and use _only_ VGA or DVI connectors to your LCD or Plasma TV monitor. I say this, because after years of experimentation, I find that this is the only way to produce 100% perfect anaglyphs. Yes, other connectors will work, like S-Video or RGB component cables, but the difference in anaglyph quality is significantly and obviously much lower. This is so important that if I ever distributed an anaglyph DVD, I would recommend this on the packaging. The difference is like "day and night". Second of all, image quality and anaglyph quality are two entirely different things. I only own cheap low end consumer cams, but with my computer I can potentially make "as good of anaglyphs" as anybody on planet earth. The only difference between a million dollar setup converted to anaglyph and my cheap cams converted to anaglyph is the image quality. The reason I say this is because people can still judge anaglyph quality no matter what the quality of the cams are, as long as they realize they are looking at poorer quality images. Now, the real meat of the subject... I am an old internet vet and learned years ago to not be offended by anything short of "I'm going to paint your house black and steal your chickens" type threats. i.e. "personal threats". IOW, you can say anything you want about how ugly I am and I just ignore it ;-) So... with that in mind, saying that there is any _easy_ way to make the same quality of anaglyphs as I make is just like telling someone who designed an airplane that it's just as easy to go out to your car, put a lawn mower blade on the front of it and add jet fuel to the gas tank, then open your doors to "act like wings" and take off and accelerate to the fastest possible rate of speed as fast as possible ;-) If you simply take the red channel from the left image and use it, which is how "traditional" anaglyphs are made, you only have a monochromatic channel, i.e., you just deleted all the other colors (green and blue) from the left image. Also, when you only use the green and blue for the right image, you have deleted the red channel from that image. This is exactly what causes severe retinal rivalry in traditional anaglyphs... and no, anaglyph glasses _cannot_ "correct" this problem. It's impossible. So, the basic solution is to somehow _not_ delete any colors. There are a number of ways to do this, and none are easy. The way I do it is includes grayscaling the left image, because an anaglyph only requires a red "channel" from it. Grayscaling the left image gives _all_ of the original colors an equivalent shade of gray. This way, none of the original colors are actually deleted when you convert that grayscaled left image to red ("monochrome red channel" and "grayscale converted to red" are _not_ the same thing). Now here's where it gets complicated... the right image of the anaglyph requires green and blue, but the same thing applies... you can't just "delete" the red channel from it like a traditional anaglyph process does for a good retinal rivalry free anaglyph. You have to somehow desaturate the red or change the hue (again, "desaturate" and "grayscale" are not identical either, but they are very similar). Now, that was just the basics... the real "art" of making a great anaglyph comes from what you do additionally after all of that (or actually "before" or "during). At this point, you make addition aesthetic choices. Examples: How do you make a red apple look good? Should I change that red shirt to brown? Can I add more saturation to all of the anaglyph to "add more color" since the shade corrections slightly washed out the overall color scheme? (answer: yes) Should I add a tint of yellow to make the skin tones look more natural?... and on and on... obviously, working with animation, film and video is much more challenging than still images, because you have to deal with the imagery on a "scene-by-scene" basis, which sometime involves some compromises... a real professional job would probably require some masking. So, yes, there might be a thousand easier ways to show anaglyphs, but if you want "perfect 3D", you are going to have to work for it ;-) P. K. Kid Non-commercial stereoscopic 3D video: (All G-Rated) http://www.puppetkites.net | ||
| Puppet Kite Kid Rating: N/A |
Let me add one thing to this that might make at least a couple of peoples' heart skip a beat... ;-) Do you know what an anaglyph filter does? It _deletes_ colors. Now, read the above post again, and this should add another dimension to the "traditional anaglyph problem"... pun intended :-) This is difficult to visualize by just thinking about it, but basically the solution involves _never_ deleting the original colors. The fact that what you end up with is a red left image and a right cyan image and you use anaglyph glasses to "cancel" (i.e. "delete") those colors is almost a paradox. What you want the glasses to do is cancel the total image, not the original colors. PKK | ||
| ronnieMervis Rating: N/A |
PKK again you are the man. your anaglyph stuff looks super. what i think we both are saying is anaglyph video looks amazing on PLASMA HDTV monitors!!!!! why don't you release a DVD of your work? do some filming get releases . put it on the market. i just tested a JVC mini HDTV-DV camcorder, with a, BensLens...... i then ran the raw anaglyph HDTV footage, through my PLASMA HDTV monitor......i was blown away!!!!!......the simple HD anaglyph footage, jumps, off the PLASMA screen!!!!!!!!! | ||
| Puppet Kite Kid Rating: N/A |
Thanks for the pat on the back, and yes, I think the future of anaglyphs could be a "bright one" :-) I would never think of making a commercial DVD of stuff made on my cheap cams, but I would _love_ to convert a professional 3D movie to anaglyph. When I watched both Spy Kids 3D and Shrek 3D, I really got the itch to "do one my way". Even if I never get that chance, I'll still enjoy sharing "home 3D videos" on the internet. It's a blast :-) Being able to do it with cheap cams is half the fun of it :-) P. K. Kid Non-commercial stereoscopic 3D video: (All G-Rated) http://www.puppetkites.net | ||
| V.S. Rating: N/A |
To PKK I have familiar with result on your page – it is looks great | ||
| Larry Elie Rating: N/A |
How do they compare to systems designed for 'normal' eye sensitivity, such as Blue/Amber (www.colorcode3d.com)? Wouldn't they take care of most of these problems up front? | ||
| Puppet Kite Kid Rating: N/A |
Not in my opinion. My honest guess is that the yellow/blue type anaglyph (like ColorCode) was primarily an attempt at finding a solution to the red/cyan retinal rivalry problem, but it doesn't solve it. AAMOF, you can "shade correct" yellow/blue anaglyphs and make them *better* than a traditional yellow/blue version, using the same principals as with shade correcting red/cyan ones. But... red/cyan anaglyphs contain *all* the colors you need, i.e., red, green and blue (RGB), and if you can produce one with no retinal rivalry, why try to improve on that? Yellow/blue versions will just give you other problematic colors, compared to red/cyan versions... same problem, different colors. Also, personally, I can't tolerate the dark blue lens needed for yellow/blue anaglyphs. It makes me feel "slightly blind in that eye". Lighten that lens to solve that problem and you simply create ghosting in that lens... catch 22. Shade corrected, red/cyan anaglyphs really can be _great_. Your only really big limitation is the color red... but using a "rust" or brown sort of hue seems to replace that color, with usually almost unnoticeable results... or you can simply totally change the color... from red to any other color that doesn't create retinal rivalry. PKK | ||
| Larry Elie Rating: N/A |
Perhaps this should be a new thread, but I'm mainly directing it to Puppet Kite Kid based on his 2 camera work. I have 2 identical Hi8 cameras. I made a bracket. I want an anaglyph for projection on a rather large screen. I COULD do all the VirtualDub3D stuff, but what would be wrong with just taking the two videos into Media Studio Pro 7 (since I have it and am familiar with it) or Premere (which I have but don't care for) as two videos, apply say a red filter to one and a blue (or green for that matter) to the other, possibly lighten each appropriately, then recombine in a multiple exposure filter. The color filtering is applied to the entire left then entire right frame, not just fields, and to all the colors of each frame, much as your applying to a grayscale filter does; filtering may be thought of as either additive or subtractive depending on how it is applied. Both fields on each camera yield information, and the result is a field-wise as well as frame-wise anaglyph. It seems too simple... but I see no reason it wouldn't work quite well.... What am I missing? | ||
| Puppet Kite Kid Rating: N/A |
The traditional color anaglyph process is actually very easy to describe. A video image is made up of three colors, red, green and blue. All you do to make a color anaglyph is make the left image red and make the right image green and blue, using any number of methods, probably including the one you mention. The red and green-blue (cyan) glasses cancel the respective colors to black, blocking the *other* eye's view. Now, that is the "traditional" anaglyph... now, here comes the bad news (or good news, if you like to work at it ;-). Some folks, including myself, have spent a very sizable amount of time perfecting the color anaglyph, mainly by coming up with advanced ways of correcting the shades to eliminate retinal rivalry. Also, recently, I have been sort of "re-saturating" the resulting colors to make the imagery more colorful and exciting. The two-edged sword, here, is that this shade-corrective and color-optimizing process is extremely complex and difficult to do perfectly, and once you experience it (or even degrees of it), it's pretty hard to go back to viewing "traditional" anaglyphs, because you become aware of all of the artifacts caused by that process. PKK | ||
| Puppet Kite Kid Rating: N/A |
BTW, the flaw in your system is that you are _first_ applying the red and cyan (or whatever) filters. Once you do this, you are unable to optimize or "correct" the original colors. For example, the red filter, the most damaging of all, totally deletes red (to white), and some other colors are affected in the same way at different degrees. Other color filters also change the original colors in a damaging way, by actually deleting some of them to a certain degree. Once you "delete" them, they are long gone, and they ain't gonna come back :-) The trick is to optimize or correct the colors _before_ you filter them... that actually makes logical sense, doesn't it? ...there really is no other way (I tried them all ;-). PKK | ||
| Larry Elie Rating: N/A |
So, over saturate, then filter? That has a less than appealing feel to it. 'Improving' the look of each picture in 2D (by saturation if you like) before filtering makes sense, but what is gone is gone from that eye in either case. And whatever filter is done to one eye (other than color) should probably be done to the other. | ||
| Puppet Kite Kid Rating: N/A |
I guess I really don't mean "over-saturate". I think "color enhancement" is more accurate. Really, what I have started to do, more and more, is "tweak" the colors of the original stereo pair before applying the anaglyph filters. I can't really technically explain exactly what is going on, here, but I basically try to change or add color tweaks that will "put more life" into the anaglyph, because the shade corrective process that I use tends to desaturate some of the colors, and the left eye is _always_ monochromatic, no matter what, so the trick is to keep the process shade corrected, but use more interesting color tweaks to compensate. You can probably see it better than I can explain it... and in fact, my anaglyph conversions are done in a WYSIWYG fashion. I am able to view the changes as I make them, then I just stop tweaking whenever I like what I see. I am experimenting with this *new* process with my recent anaglyph uploads. Compare the anaglyph colors to the original parallel pairs. See what I mean? You can *see* it, but it's quite difficult to totally understand it: http://www.puppetkites.net/ Using VirtualDub, I have been stacking three separate "Hue/Saturation/Intensity" filters that tackle groups of colors, separately. I also add the "Color Equalizer" last (actually called the "Color Balancer" in the filter window), which brightens the red side a bit. These screenshots just show the color tweaks... the rest of the process is also done after this (basically, the one I have on my VirtualDub3D page: http://www.puppetkites.net/virtualdub3d.htm ), except for the Color Balancer, which I always apply dead last. It simply brightens the finished red side a bit. Here's screenshots of what I used for "Giant's Eyes". I only used hue changes for this one... I just liked what I saw. Sometimes I also play with saturation and intensity, especially when dealing with lots of red in the original: http://www.puppetkites.net/vdubanaglyphtemp/huefilter1.gif http://www.puppetkites.net/vdubanaglyphtemp/huefilter2.gif http://www.puppetkites.net/vdubanaglyphtemp/huefilter3.gif http://www.puppetkites.net/vdubanaglyphtemp/colorequalizerfilter.gif PKK | ||
| Larry Elie Rating: N/A |
I've been to all your pages (before these posts), and they did indeed get me thinking about it. Look, 30 years ago I was doing physical optics, so I'm not fresh, but my background is physics, so here's they way I see what you have done. The red filter is a modestly broad band Gausian filter probably tapering badly. To do it right requires a 'notch' filter, but that would probably cost too much in quantity, so we use the cheap filters. When I get a chance I will scan one in a film scanner to get the attenuation by frequency. What you have done is to increase the saturation (or intensity) of everything that the red filter would attenuate so that much of it will still come through afterward. This way the image stays 'bright'. Ditto the other color. What I'm proposing is to essentially do a 'positive' filter; knowing the attenuation by frequency, goose up the other colors for the red and then for the blue on each of the frames and fields for that eye. You leave just enough difference to get good 3D. You won't lose anything. It makes the cheap plastic glasses seem like a notch filter. It should be pretty much the same as what you have done, just automatic. I'm going to work out the math (well, I'm on vacation next week) but this should not have to be art, just simple math. I'm not being critical at all, but I think this can be scripted and then would be straightforward. Let me try it and get back with you. | ||
| Peter Wimmer Rating: N/A |
Although I have scientific background, too, I don't believe that solving anaglyph problems just using maths is the proper way to do it. During my work on anaglpyh algorithms, I found out that non of the mathematically deduced algorithm works better than results done by manually tweaking some parameters. Most important for good anaglpyh performance are the choice of the display and the glasses to avoid ghosting. Andrew Woods from Curtin University of Technology wrote a excellent paper on this topic, see http://www.curtin.edu.au/cmst/publicat/2004-08.pdf. Another important issue is avoiding retinal rivalry. They best way this is solved by using my optimized anaglyph algorithm: http://mitglied.lycos.de/stereo3d/anaglyphcomparison.htm. For my Stereoscopic Player (http://mitglied.lycos.de/stereo3d), I improved the optimized anaglyph method so that the red channel information doesn't get lost but mapped to green and blue channel instead. Currently, this algorithm is only available in the development mode of my player, however you can activate it by pressing Ctrl+Alt+D. Ctrl+Alt+A allows to tweak the algorithm. Larry, most important for you 3d movie is that you don't create an anagpyh version only, but another version in side-by-side or over/under format too. Even if you don't have shutterglasses or a polarized projection system at the moment, you might buy one later or want to show your movie at stereoscopic meetings. An anaglyph movie, however, can't be used for such systems. | ||
| Larry Elie Rating: N/A |
Thanks for the paper link; I was going to use a negative scanner and try and figure out what the transmission of the glasses was, and pre-compensate. Now I can just use the transmission data, which saves half the work. First off, my simple project has grown. I have built a bracket using 2 identical Hi8 cameras I happen to own. I'm planning on using a significantly larger than normal stereo base seperation as I want differences to be rather dramatic, and most things I photograph will be 10M + away. I am syncing the cameras by firing a flash after starting each, then shifting the video to match so I don't need to worry about GenLock; both cameras are clocked by a quartz oscillator so this should be good enough. I will save the entire project including the original tapes so any combination will be available for comparison. The projector I have is an Proxima 9270, a 3 LCD 3500 lumen projector that is not in the article. Since the color is dependent on the filters on the lcd's which are in turn chosen based the the color temperature of the light source, my system probably better matches one of the VT's. Unfortunately, since I have to buy a substantial number of glasses, the SpyKids ones are probably my only choice, but the cross talk is lower than I expected. Again, one should be able to reduce cross-talk if you know the system inputs and responces. As far as using 'only math' to do art... well, let me put it a different way. I take about 3000 digital pictures (99.44% 2D) a year. I pass 95% through Paint Shop Pro 8.1's "Enhanse Photo", for artistic reasons (exceptions; sunsets, and some night shots with color lights) but the art here is simple math; a color balance, a contrast enhansement (re-distribution of entensity), a few smoothing filters and saturation adjustment. It looks like art, but it's math. What I want to do is an "Enhanse Anaglyph" and let it do each field of each scene be fixed as lighting and other parameters changes on it's own. This is most important because the hardest things to get 'good' with 3d anaglyphs (ghosting excluded) are DARK. This is where you pretty much already know what you want. It will take plenty of computer time, but not as much of my time. I don't think we are disagreeing all that much on this. I'm going to try and do some day and night parades (motion and lots of tricky colors), and see what I get. | ||
| Puppet Kite Kid Rating: N/A |
Here's a little more advice to hopefully save you some time experimenting... I have "walked many miles in these shoes" the hard way, so this might help you to not to repeat my mistakes :-) First of all, concerning anaglyphs, "color correction" to eliminate retinal rivalry and "color enhancement" to make the overall color scheme more vibrant and exciting are two different beasts. You can use mathematical formulas or algorithms to equalize the shades, but you can't use "math" to make aesthetic choices that change scene to scene and mood to mood. Also, when I use a set of filters for one scene, many times it won't even _work_ on another scene... period... all sorts of strange artifacts can show up if you try to use one color enhancement for all scenes. This stuff is much harder to do well than many people might think (I'm still learning :-). Consider that still image anaglyph fanatics spend hours on "one piece of art". We have 30 of those "pieces of art" every second to deal with... sometimes nearly impossible to work out, and sometimes you just have to quit because it just won't work. Now, about syncing cams: If you don't somehow "sync" your NTSC cams, the best sync you can get during post processing is "no worse than" 1/120th of a second... and this is only if you are willing to separate the fields to 60fps to find the closest two "common" frames. For many moving subjects (including common things like waving hands or cars going down the road), this will not be a good enough sync. Trust me... I spent about 3 years working with non-synced cams, and I used many tricks to work around the sync problem... motion blur via frame blending works the best, but still isn't perfect. For the past few months, I have been monitoring my degree of sync, and I turn one cam off and on until I get a "good enough" sync for my project. Shooting a turtle race might only require a 1/60th of a second sync (no syncing needed) or a movie of a car race might require a 1/2,000 of a second sync to be perfect. You will also be limited on the amount of time your cams will remain in a certain degree of sync. For example, mine can retain "better than" a 1/500th of a second sync for about 20 minutes if I start them where the sync is nearly 1/500th of a second, and the sync is getting _better_. It takes 10 minutes for them to "drift" into perfect sync and 10 minutes for them to drift back apart to 1/500th of a second. I can monitor this as I shoot with a tiny voltage gadget that plugs into my "A/V" outputs. PKK | ||
| Larry Elie Rating: N/A |
Yep, I understand the sync issues; I just don't want to take the time to sync the gate times (shutter speeds) to make it work. Happily, my cameras are indeed both quartz gated, and right side up (although with CCD's that isn't a problem... with vidicons that would be a show stopper) but I'm not filming anything faster than a walking pace... for now. Hopefully, I can even get away with frame wise as opposed to field wise stuff at these speeds. I'll know better in a week or two. |
