Hi there... I recently decided to embark on a monstrously large project: I am making an open-source content creation system. (similar to the unreal content system, but nowhere near as complex) This system will be designed to be a content creation system tailored to image-based 3D graphics and lighting, for photorealistic graphics. (its some pretty amazing stuff) I am pretty well-researched in this field, so I think this project is doable for me. However, I have run into a wall... One step in my content pipeline is a tool that I just cannot get right. It is definately non-trivial, and although it certanly is possible, it is hard (at least from my perspective.) I am looking for interested programmers to help me in making an open source tool that may become a very useful part of a highly advanced graphics system (you should definatly check out the cutting edge image-based research. really sweet. The demos and theories are jaw dropping) Anyhow, to get back to the problem; I am trying to develop a command-line program that will discover (and correct maybe) the first two radial distortion coefficients in photographs. There are many methods of accomplishing this, but I am not sure I understand the calculus involved in most of them so I really think I will need some help. To be honest, any solution would be a good one, and the least required user interaction required the better. just to get started, a google search on radial distortion correction is a good placce to learn, as well as Camera Calibration and Paul Debevec PH.D thesis, chapter 4 Is there anyone willing to help me tackle this problem? It would mean more useful software under the GPL, and a possible huge advancement in the graphics community. Or at least it might be fun to solve a new problem for all you gurus out there tired of answering the same old questions on this forum. It would help me out a lot. Please respond if you are interested, and In the meantime I will start wrtiing a framework and base code to start from, and anyone who wants to help can add on to it with me. thanks

Quote: Original post by Steve132
<cut>
This system will be designed to be a content creation system tailored to image-based 3D graphics and lighting, for photorealistic graphics. (its some pretty amazing stuff)
<cut>

(you should definatly check out the cutting edge image-based research. really sweet. The demos and theories are jaw dropping)

An URL would be interresting here?

Quote:
Anyhow, to get back to the problem; I am trying to develop a command-line program that will discover (and correct maybe) the first two radial distortion coefficients in photographs.

After discovering what this is and why I'm probably not the one to write such a piece of code for you, I'd like to ask you, why on earth would you need this in a content creation system? Isn't the point of a content creation system that you create (amongs others) graphics by techniques like rendering and so don't need to correct camera-lens errors?

Quote:
<cut>
Is there anyone willing to help me tackle this problem?
<cut>

Willing: yes, capable: no. Anyway, good luck!

Quote: An URL would be interresting here?

I am a big fan of THIS
also, google Image Based Lighting to find some very neat videos.

Quote:
After discovering what this is and why I'm probably not the one to write such a piece of code for you, I'd like to ask you, why on earth would you need this in a content creation system?

Well, first off, I would like to clarify: I don't want anyone to write the code FOR me. I am simply looking for people who would like to join brains with me so to speak, so we can work together to understand and solve this problem.

Quote: Isn't the point of a content creation system that you create (amongs others) graphics by techniques like rendering and so don't need to correct camera-lens errors?

Yes, yes it is. This system will be a normal content creation system in the respect that it will be used to create graphics and rendering. However, the system I am working on will be designed to incorporate Image-Based processes and methods of creating those models. For example, I plan to have a High Dynamic Range image processing tool for lightmaps and HDR textures from normal photographs. I also plan to have a world creation system and script engine similar to the unreal engine, among several other cool features.

However, perhaps most important would be the image based modeler, which uses photographs of real objects to interpolate their geometry, and also to extract camera positions, and use those positions to reproject the geometric models into texture coordinates. These new, photographically created renderable models could then be used in 3D applications in the same way normal textured models are, except they would have a very high degree of realism. (Just look at the Campanile)

All of these things are possible, but it is crucial for any photographic operations that the photographic radial distortion of a source image (and ideally some other parameters) be corrected, so that the image represents the image one would get from a perfect linear perspective model, like the one in OpenGL (such perfection is impossible to achieve with actual lenses)

Now you understand, I hope, my master plan :)