This is a counter-view on Synthonics' rebuttal to my email forwarded by Bob. This is just a personal opinion and does not reflect the opinion of my company. Warning. These statements are full of personal opinion, conjecture, and forward looking statements, etc. etc.
I agree with Synthonics' that photogrammetry is the appropriate digitization (i.e. 3D scanning) method for some applications. It is usually the method of choice when it is not desired or difficult to place the object in front of a laser scanner or mechanical arm. For example, laser scanning a huge building is probably not as efficient (if even possible) as taking some photos and using a photogrammetric software product. Also, I agree with Synthonics that the low-cost of photogrammetric software on a PC, coupled with affordable digital cameras and flat-bed scanners, make photogrammetry a 3D scanning technology accessible to individuals and small companies that can't afford a laser scanner. Especially attractive is that photogrammetry can produce very realistic objects because the photo texture of the object is captured simultaneously with the geometry. However, the tedious nature of the method means although it is possible, in my personal opinion, it is not likely on the scale that Synthonics' capitalization would require.
In fact, most of these applications are already validated by the companies EOS (www.photomodeler.com) and 3D Construction (www.3dconstruction.com). Both of these companies ship now photogrammetric software for the PC which is being used *NOW* in projects of the types espoused and proposed by Sythonics. Go to the web sites (www.3dconstruction.com and www.photomodeler.com) and see real completed projects from a myriad of users, including museum artifacts! Compare this to the "vaporware" pronouncements of Synthonics. On the one hand, you have these photogrammtric companies with profitable software and then you have Sythonics with their inferior Wireframe Express product as a sink for capital piling up patents. Hmmm. Makes me scared of their accumen for legal manuevers, so I made sure I did my homework. By the way, I am not a representative of these photogrammetric companies, although I have used their products. I have real experience using photogrammetry and have done projects ranging from a few hundred up to a couple of thousand 3D points.
As for "File sizes small enough for easy Internet transfer" and "Accurate portrayal of artifacts in three dimensions", the MetaStream technology is incredible in this area. It not only defines extremely small files using a patented color + geometry encoding (from famed Russian scientist, go to www.real3d.com), but it is also a "streaming" (i.e. progressive) format technology. This means we can start seeing the 3D object before the whole file has been tranfer, much how GIF graphics on the web "window pane" in as they transfer. Synthonics hasn't announced any super technology in the compression and transfer of 3D data. They seem to just be refering to the low polygon counts that photogrammetry is best suited for. Note, that recent software advances make creating such low-polygon count models from laser scans nearly if not automatic!
What I am saying in nutshell is:
1. SMALL MARKET SIZE: Photogrammetry will not be a huge market, at least the Synthonics' methods I have seen thus far in awarded patents. (Note, no way for me to know about things they have not released publicly.) Even the potentially mainstream application of individuals and small business will not pan out in my opinion because, except for box-shaped objects, the time and tedium for user to mark a few hundred points in several photos (for 360 degree view) is more than mainstream markets have patience for. In my opinion, mainstream will wait for laser scanners to be installed in their local Kinkos Copier store. It is like arguing that Xerox machines are too expensive, thus everyone should use a sketch pad. Dr. Palm seems to have some past experience in copying machines.
Let me add that photogrammetry as a supplement to higher order 3D morphing (something which no one owns because it is prior art in public domain) does have the potential to be a mainstream product as will be evident in our 3D human face software product. In fact, we don't even use photogrammetry in traditional sense in our product. By the way, the Synthonics patent does mention simple 3D morphing (linear interpolation of vertices) but it does not make any claim on this. They have no patent on 3D morphing. Their patent is on using photogrammetry to find the positions of vertices of an existing mesh. This means each vertex, one-by-one.
2. WEAK TECHNOLOGY: For the applications that photogrammetry excels in, the companies with the best products and technology may win. In this case, EOS and 3D Construction seem to me to be way ahead of Synthonics in several areas:
a) accuracy of 3D geometries
b) flexibility to use any camera
c) flexibility to use 3 known points or other combinations for orientation
d) a global (bundle) statistical optimization that refines the orientation to provide very accurate solutions. This provides better registration of photo texture maps among other things.
e) perspective correction on photo texture gets rid of warping and misalignment errors
There are more details but you all might get saturated.
3. LOTS OF COMPETITION FROM EVERY ANGLE: Not only is Sythonics threatened by more established photogrammetric players with better technology, but Synthonics shoots (and sets the size of their capitalization) to go after markets of the "big boys" like MetaStream (Intel/Microsoft/Lockhead involvement), Viewpoint, laser scanners, etc..
Synthonics' press releases talk about creating any 3D object easily and HDTV. These are pipe dreams in my opinion. In my opinion, it will not happen using their what I understand to be their inferior photogrammetric technology as stated in their patents. Will probably never happen with even the best photogrammetry technology. Photogrammetry will remain for specialized application, which Synthonics' seems to admit somewhat in their rebuttal to my email.
To the best of my knowledge, today there is no published photogrammetric technology that can make any 3D object desired with a modicum of user work load. There are specialized objects and environments where photogrammetry is best, even though it still takes an tremendous amount of user work.
Although automatic feature extraction, shape from shading, shape from stereo, deformable templates, and other research in object reconstruction is promising, it is far from being commercial technologies and besides these are not the basis of Sythonics' technologies or patents, as far as I know.
4. POOR FINANCIAL POSITION: When it becomes clear to Synthonics (well I should say to it's investors) that photogrammetry is not a mainstream technology, but a niche technology for specialized applications, then the apparently relatively (compared to apparently private held photogrammetric companies Eos and 3D Construction) huge capital base that Synthonics has a liability for (and has apparently already spent according financial documents referenced and statements in this thread) will be a weakness in their attempt to compete in these smaller markets. In other words, they sold themselves big and they will need to find significant revenues from small markets.
I am impressed that Synthonics could sell Smithsonian on their technology, if that is really true. We have to see something tangible on balance sheet. They do seem to have very good (short term) marketing :-). I don't fall for it, but it seems 3D is so new and misunderstood that many do.
******************************
Since some of you may doubt my assertions about Synthonics' patents, I will provide proof of prior art on one of their patent applications that doesn't affect me. I must hold secret the proof of prior art on others, since it is a war chest against any future Sythonics action. I very worried about what will happen if Sythonics starts to go bust and they turn to their patents and legal maneuvers.
In the patent "Method and Apparatus for Creation of 3D Wireframes", Synthonics refers to another patent _application_ "Method and Apparatus for Image Correlation for Three Dimensional Image Production". As far as I know, this patent has not yet been awarded. They give a brief overview in the aforementioned _awarded_ patent. Basically this latter patent application appears to be about projecting a light pattern on an object in order to achieve "pretargets" (i.e. reference points) that can either be marked by a user on two or more photos, or perhaps automatically correlated by algorithm. Basically this method is important for cases where you don't want to put marks on an object and there are not enough natural marks on the object, i.e. museum archiving (Smithsonian?).
This patents was applied for on Oct. 4, 1994. I found a research paper which describes essentially the same invention and was published in 1992. Thus, the Sythonics patent, if ever awarded, could be overturned in court unless Dr. Palm could prove he invented earlier or his invention was unique (note I am not a lawyer):
[21] Saji H, Hioki H , Shinagawa Y, Yoshida K, Kunii T (1992) Extraction of 3D shapes from the moving human face using lighting switch photometry, eds. N. M. Thalmann and D. Thalmann, Creating and Animating the Virtual World (Springer-Verlag) pp. 69-86
|
