mattbjerregaard
08-27-2008, 10:50 AM
Hi all,
Intro: I am an editor/compositor and camera director primairily but am currently involved in helping out voluntarily in an underwater conservation project which is unfunded and involves the monitoring of a specific extremely important coral reef in the Philippines. I am also a professional diver with many years experience.
What I am trying to achieve is a massive texture stitch/record of an area of coral. My camera rig consists of a HDV camera with levels mounted so as its Z axis is as fixed as possible, perpendicular to the sea bed, and Y and X remain as level possible also. I have rigged a trio of crossing lasers to give me a fixed distance from the sea bed. Basically I swim or use a propulsion vehicle over a massive area, and record the corals on the bottom in matching, 'stitchable' stripes/strips.
So far I need to repeat this process about every month for the entire area (about 2 square km, it is massive) and I am manually sitching the frames together which takes far too long. This is incredibly time consuming and very inaccurate. The portion of the image I am concerned with is the centre strip. I need to automate the process more... I would look at buying a copy of 2d3 Tacitview but it is $$$
I have had a few ideas... They are probably crazy... Maybe I am barking up the wrong tree but I need to create in Maya something like a slitscan technique (Zbig et al) and bake out the texture.
Basically I can remove lens distortion and what I need to do I think is track the camera in something like Boujou then take the data and bake it to a single surface. I don't know how to do this for a texture using multiple frames from a camera projection.
Or in maya live, track a frame, solve it, bake the centre strip, repeat until the image sequence is completed. In effect you are baking the film back not the sea bed... headache
CONCLUSION
Any offers extremely grateful.
The end product does not need to be at an FX level of realism, the data just needs to be there.
Please excuse my ignorance and remember that this is for an extremely worthy cause.
True environment mapping.
Many thanks for reading...
Intro: I am an editor/compositor and camera director primairily but am currently involved in helping out voluntarily in an underwater conservation project which is unfunded and involves the monitoring of a specific extremely important coral reef in the Philippines. I am also a professional diver with many years experience.
What I am trying to achieve is a massive texture stitch/record of an area of coral. My camera rig consists of a HDV camera with levels mounted so as its Z axis is as fixed as possible, perpendicular to the sea bed, and Y and X remain as level possible also. I have rigged a trio of crossing lasers to give me a fixed distance from the sea bed. Basically I swim or use a propulsion vehicle over a massive area, and record the corals on the bottom in matching, 'stitchable' stripes/strips.
So far I need to repeat this process about every month for the entire area (about 2 square km, it is massive) and I am manually sitching the frames together which takes far too long. This is incredibly time consuming and very inaccurate. The portion of the image I am concerned with is the centre strip. I need to automate the process more... I would look at buying a copy of 2d3 Tacitview but it is $$$
I have had a few ideas... They are probably crazy... Maybe I am barking up the wrong tree but I need to create in Maya something like a slitscan technique (Zbig et al) and bake out the texture.
Basically I can remove lens distortion and what I need to do I think is track the camera in something like Boujou then take the data and bake it to a single surface. I don't know how to do this for a texture using multiple frames from a camera projection.
Or in maya live, track a frame, solve it, bake the centre strip, repeat until the image sequence is completed. In effect you are baking the film back not the sea bed... headache
CONCLUSION
Any offers extremely grateful.
The end product does not need to be at an FX level of realism, the data just needs to be there.
Please excuse my ignorance and remember that this is for an extremely worthy cause.
True environment mapping.
Many thanks for reading...