| Author | Message | ||
     Michelle N (She11z) New member Username: She11z Post Number: 1 Registered: 2-2011 Rating: N/A |
Hello, I have recently purchased Facegen to create 3D models of faces for a thesis im currently working on. As Facegen does not do stereoscopic images, I need to export the single face image (.bmp) to another software program. My question is - does anyone know of any software programs that can allow me to create a stereoscopic representation of a image from facegen (just one image). All the programs I have looked at so far require you to import two images to create a stereoscopic view (a left-eye view and a right-eye view at about 6.5cm apart). As my face model is an animation, I need to find a software program that allows me to create a stereoscopic view from only one image (so the program needs to be able to artifically create this left-eye/right-eye view for me). Does anyone know of any good programs? Cheap ones too would be good! Thanks, Michelle | ||
| Christoph Bungert (Admin) Board Administrator Username: Admin Post Number: 56 Registered: 4-2006  Rating: N/A |
Stereoscopy requires 2 different point-of-views. This is the whole principle of it. By artificially creating a stereo-pair from a single image you're entering the realm of 2D-to-3D conversion. I won't recommend to go into this direction. Couldn't you render two camera positions in your software? Christoph | ||
| Michelle N (She11z) New member Username: She11z Post Number: 2 Registered: 2-2011 Rating: N/A |
Hi Christoph, Facegen doesn't allow you to numerically render 2 camera positions, so im not sure what to do. I realise that 2 different points of views are required to create a stereoscopic image - and I do have a 3d camera + everything necessary to achieve this - but I don't want to take photos of real faces. I want to use 3d models of faces from facegen(to overcome some limitations in face recognition research). Thus, I need to be able to create a stereo image of a model, not a photo. Do you know how I could achieve this? Also, why don't you suggest 2D to 3D conversion is a good idea? I am quite new to this, and so any advise is appreciated! Thanks, Michelle | ||
| Michelle N (She11z) New member Username: She11z Post Number: 3 Registered: 2-2011 Rating: N/A |
Also - I can get access to photoshop at uni, which I understand allows you to create stereo images. All I really need to do then is to create a second image of my face image that is displaced approx 6.5cm to the left or right. I can then export these 2 pics into photoshop and create a stereo image. I'm just not sure how to create a slightly displaced 2nd image.. any thoughts? | ||
| Christoph Bungert (Admin) Board Administrator Username: Admin Post Number: 57 Registered: 4-2006 Rating: N/A |
If you create a stereo-pair, by using the same image and just adding a displacement (parallax) what will be the result? Well, the whole image will be behind (or in front) of the monitor plane. The image in itself however will remain completely flat. The depth information in a real stereo-pair comes from the variety in the displacement. The challenge in doing 2D-to-3D conversion is giving every pixel the right displacement (parallax), which is from a theoretical viewpoint almost impossible. The original depth-information is missing. Via conversion worthwile results are possible, but they're never the real deal. Conversion means a lot of effort. The Hollywood studios are currently investing about 20-30 Million Dollars per movie to convert their 2D movies to 3D and IMHO it's not worth it. (For example Clash of the Titans, The last Airbender, Green Hornet). Christoph | ||
| Michelle N (She11z) New member Username: She11z Post Number: 4 Registered: 2-2011 Rating: N/A |
Thanks for all your feedback Christoph! Its been very helpful  Michelle | ||
| Larry Elie (Ldeliecomcastnet) Senior Member Username: Ldeliecomcastnet Post Number: 144 Registered: 10-2006 Rating: N/A |
The requirement is a DEPTH MAP. It can be generated by the user. It is a lot of effort; just like any serious photo editing. There ARE applications that try to do it automatically, or by hand. Of course by hand is better. An example is '3D Combine'. The auto works to some degree on simple depth (scenery) but if you want to actually generate the depth map completely, it's a lot of work. Another example of one that can work if you put forth the effort is 'Photoshop CS5'; you can find a how to video online. Given effort, you can do it very well. Often, you just find one color channel that is close enough to work and go with it; sort of like picking out one picture from a batch. The reason the studios are working hard is that there are so many frames involved and you can't pick just one frame of many! The scenery is easy; the close-ups are hard. Before you say that isn't good enough; most lenticular pictures you buy are only 3 or 4 levels, and that isn't hard at all. The studios are gettting that for a few million. CONTINUOUS depth is hard. That would be very expensive. | ||
| Michelle N (She11z) New member Username: She11z Post Number: 5 Registered: 2-2011 Rating: N/A |
Hi Larry & Christoph, Thanks for all your feedback - it's been very useful. I passed on your messages to my department supervisor,and he wrote back the following message. Are you able to advise accordingly? - "What the stereo forum folk told you is absolutely correct. But all of this relies on an original assumption - that we do not have 2 images of our facegen model. The challenge here seems to be to get hold of these 2 images. The whole point of facegen, as I understand it, is that you create a whole 3D model of a head in the computer's code/memory. This means that you can turn the head around to different pose angles (pitch, yaw and roll), etc. The 3D head files can be saved and imported into other programs, and retain their whole structure. To get our 2 images of the face/head, all we need to do is make sure we get hold of an image from one viewpoint (e.g. a photo from one viewpoint), and another photo from a slightly different viewpoint about 6.5cm to the left. These 2 images will then form a stereo pair, so that we can show it on our cool stereo monitor. So if facegen can create the 3D model of the head, and save it in a file, then we just need a program that has the facility to open these files, and then take the "photos", and save them as 2 image files (.jpg or similar). Do you know what software can achieve this?" Thanks, Michelle | ||
| Harvey Jewett (Harveyfloyd) New member Username: Harveyfloyd Post Number: 3 Registered: 3-2011 Rating: N/A |
You should be able to save your models out of Facegen with the .obj extension and open them in Photoshop CS3-5. You can then render the face from any angle you like and save left and right views or create an anaglyph or even a lenticular print. You could also open the file in any number of 3D modeling programs from open source Blender to DAZ Studio to Studio Max. |
