Microsoft Photsynth: 2-3 pics transformed to 3D

The current BBC news has an article "Photos transformed into 3D model" in todays Technology News.

I am going to be able to finally convert my many sets of snaps of great sculptures to a virtual reproduction that I can zoom in on, rotate, even fly through.

This is not the first software to turn two-dimensional images into 3D models. However, until now the software has had limited rendering or else is costly.

The new Microsoft software, to be presented this week this week at Siggraph 2006 in Boston.

It needs as few as just 2-3 pictures of an object and by matching common features, builds a 3D model without needing a turntable or laser set up for carefully sampled views.

"Technology that transforms digital images into 3D models will be unveiled at a conference on Wednesday.
Microsoft's Photosynth takes collections of images, analyses them for similarities, and then displays them in a reconstructed 3D space.

The system, to be previewed at a computer graphics meeting in Boston, will allow users to walk or fly through a scene to see photos from any angle.

Microsoft says Photosynth should be available for use later this year."

"Microsoft have said they believe the technology will almost certainly be web-based, and people should be able to run the application later this year"

So how could you use this software?

Asher

Hmm, not sure what's so new about this. I did some 3D prints a while back using some free software I downloaded. It worked so well you can almost see the effect in a flat photo I took of the print. As you walk past the print viewpoint shifts and you can see behind stuff etc. Admittedly this wasn't done with just two or three images, but iirc you could get by with that few as the software would interpolate if you didn't provide enough frames.

Probably what's different is that it's modeling based (CGI) rather than optical.

- DL

I wonder whether MS has purchased, ie gobbled up some small existing software company.

In any case, it should be interesting.

Asher

Asher Kelman said:

I wonder whether MS has purchased, ie gobbled up some small existing software company.

Click to expand...

Have they ever done anything else? Even original MSDOS was bought from Seattle Computer. I think Basic was the only thing written from scratch, and I'm not so sure about that

- DL

Well, iview is now part of Microsoft. I hope they'll make it more robust? Do they fund the current teams or have they a record of using their own programmers?

Asher

Hi Asher,

thanks for the post. I have spent a fair bit of time, on and off, looking at this sort of thing. I think, as usual, M$ are telling porkies. Having said that, it is feasible using a few photos of selected objects from selected positions.

A photo is 2d, an object is 3d. There just is not enough 'space' to contain the depth info, unless you sacrifice something, say colour. I have some cnc software which can be used to generate the code to machine a picture, darker shades deeper, or vice versa. As an excercise, if anyone can easily/automatically produce a jpeg of a human face, say, with distance from the lens represented by a shade of grey, then I would be very interested to know how....

Consider a simpler problem, that of representing a straight line by a point - from 2d to 1d - use colour of point to represent length of line, say, but what do you use for angle??

At the moment, there is no software that can even convert a raster image to vector - e.g. jpeg to dxf format, that is reliable/low cost/automatic. I think M$ are doing something slightly different than what I want (I won't add 'as usual')

Best wishes,

Ray

(afaik porkies= pork pies= lies)

Ray,

Since you have already dabbled in 3D, may be more than that, tell me what would you use, preferably on a Mac, to transform say 10 handheld pictures around a sculpture to 3D. I want to work with this to practice.

After that I'll make a turntable and do things more exactly.

Asher

P.S. it can take any time to render, I can dedicate a computer for that!

Hi Asher,

About Mac's, I know nothing. On a pc I've used various low cost software /free/written bits of code, manual measurements, etc.to do panoramas, anaglyphs and the like, but your requirement is inside out, so to speak, you will need to eliminate the background. (for panorama's, you sort of need to eliminate the foreground.)

I think you could, with care, initially stitch with photoshop, but then you would not be able to get continuous rotation, if that is what you want, and I can see numerous problems arising with the stitching. I suspect it would be one good use of a scanning back!

I think there are possibly some animated gif preparation software products that may do it, I need to check back, but as mentioned mine was all PC based sw.

Best wishes,

Ray

More info on this page:
http://labs.live.com/photosynth/

Thans for the link, Bart.

I watched the video, only a small amount was factual. However, be aware of microsoft's intent with this.

What they are proposing is not something that I see as very clever, and is of little real use, although ymmv.

It is nothing along the lines of what I was hoping it would be. It is a great idea for generating a requirement for more bloated os's and hardware.

The bbc very rarely gets any technical news reported correctly, but I suspect its better than most.

Best wishes,

Ray