| Author | Message | ||
     lisa lee Rating: N/A |
Anyone have any experience with various 2D to 3D converters that are available? Do they really work? And if so does anyone know of any reasonably priced converters. | ||
| Giorgio Bogoni Rating: N/A |
Does anyone know where I can buy a black-and-white to full-color automatic movie converter? :-) Sorry lisa, there's no depth info in 2D movies ... and no way to reasonably generate it. Anyway there's a big business around this kind of software, I think this is why we are going on speaking about this. Giorgio. | ||
| Anonymous Rating: N/A |
It works O.K. only for specific (horizontal) camera motion. No good solution for general case. | ||
| Anonymous Rating: N/A |
Yeah there are movie converters. I even downloaded a freeware 2d to 3d converter on the fly. Its called movieplus, I found it just by searching www.google.com for different things... like stereoscopic movie convert stuff like that. Try adding movieplus to the search and I'm sure you could find it. It has a bunch of options (play in 2D, eyeFX format 3d, 2d to 3d general conversion, and 3d to 2d general conversion) and settings for above/below, relaxed or cross eye and interlaced. It also lets you control the 3d effect (seperation) so you could find the best setting for 3d but avoid ghosting. Like the last post said, not as good as an 'original' 3d, but great for all the mpg, or avi files you have laying around... Hope this helps... Justin | ||
| Richard Scullion Rating: N/A |
The new release of 3DCombine adds a 2D->3D converter. You have to create a depth map to tell it what goes where though. There are instructions and a sample in the help file. www.3dcombine.com | ||
| caused Rating: N/A |
Richard, That sounds cool, what will you do about the missing information? Stretch, clamp, or blacken/alpha out? It would be nice to have those kinds of options, so that either automaticly fix it, or be able to paint it, or even possibly in photoshop use layers, to combine different images to get the missing information. | ||
| Marvio Rating: N/A |
Richard, I tried that 3d converter that you mentioned and the only thing it did was to sink everything into the screen? I mean all objects were sank. In fact, the only way to tell it's working at all is to have your windows task bar at the bottom of the screen, then you'll be able to notice that all objects are futher then that, but once the task bar is gone, you loose any sense of depth, becasuse all objects have the same "sank" depth? Am I missing something here? | ||
| Richard Scullion Rating: N/A |
If you want proper 2D->3D conversion you need to create a depth map and use the manual 2D->3D conversion, not the automatic one. There's information and samples in the help file. The depth cueing automatic mode does just drop the image back into the screen - but it never claims to do anything else. The other automatic modes are much cleverer but make certain assumptions about the picture you are converting which may or may not be true. | ||
| Richard Scullion Rating: N/A |
Caused: The algorithm for replacing the missing background information is fairly complex and uses a range of techniques. I think it does a fairly good job but there's nothing to stop you pulling the picture into a paint program to touch up any areas you aren't happy with afterwards. | ||
| Anonymous Rating: N/A |
The only really good 2D - 3D conversion I've seen is that which involves human beings in the loop to perform a process analagous to colorizing. DDD, Another World, and a few others have converted video clips using this technique that actually look quite impressive. As I understand it, the human operator uses his own knowledge of what the shapes of things ought to be, what's in front and what's behind etc. to create a depth map for the image. The software then basically shifts and distorts the forground and background objects to create a pseudo stereo pair. You can see what's happening if you look closely - parts of the background tend to get stretched to fill in the space created when the foreground objects are displaced. However these companies only perform their services in house - they don't sell their software. The so called real time "2D-3D converters" that I've seen (and I think I've seen most of of them) are all junk in my opinion. First of all, the cheap ones do nothing except displace two copies of the same image so that all you get is a flat image that looks like it is behind or in front of the screen. The more sophisticated ones try to determine about what is in front and in back by analysing motion and (in at least one case) things like contrast and focus. Then they try to do the placements and stretching automatically. None of them do it at all well, IMHO. You constantly get wierd stereo, which reverses and shifts in parallax constantly. I captured frames from one once and on even casual inspection they looked positively weird. Some will work fairly well for *very* specific types of clips, and these are what are usually shown as demos. For 99% of the video in the world, though, they suck. | ||
| Anonymous Rating: N/A |
The only really good 2D - 3D conversion I've seen is that which involves human beings in the loop to perform a process analagous to colorizing. DDD, Another World, and a few others have converted video clips using this technique that actually look quite impressive. As I understand it, the human operator uses his own knowledge of what the shapes of things ought to be, what's in front and what's behind etc. to create a depth map for the image. The software then basically shifts and distorts the forground and background objects to create a pseudo stereo pair. You can see what's happening if you look closely - parts of the background tend to get stretched to fill in the space created when the foreground objects are displaced. However these companies only perform their services in house - they don't sell their software. The so called real time "2D-3D converters" that I've seen (and I think I've seen most of of them) are all junk in my opinion. First of all, the cheap ones do nothing except displace two copies of the same image so that all you get is a flat image that looks like it is behind or in front of the screen. The more sophisticated ones try to determine about what is in front and in back by analysing motion and (in at least one case) things like contrast and focus. Then they try to do the placements and stretching automatically. None of them do it at all well, IMHO. You constantly get wierd stereo, which reverses and shifts in parallax constantly. I captured frames from one once and on even casual inspection they looked positively weird. Some will work fairly well for *very* specific types of clips, and these are what are usually shown as demos. For 99% of the video in the world, though, they suck. | ||
| Anonymous Rating: N/A |
some of the video ones don't analyze motion at all, they just delay one of the fields. |
