Camera distortion

[latrommi_suineg]

im working on a project that incorporates live (blue screen) footage into a 3d environ... the tracked images( camera outputted by tracking software) have real world camera attributes and distortions that i can never seem to achieve using the maya cameras on their own... besides having constraints to impede tweaks from cancelling the effect of each other out, there are focal depth/ rotations timed together which create the look a real world camera can give...
without the data to input already tuned, can someone offer me a formula or a resource whereby i may achieve more realistic camera views... i remember working in softimage and a perp cam was not to be tweaked or zoomed as its proportions were so easily disturbed... but even in a scene with camera shaders and focal data entered the maya cam still looks like an orthographic cam with z translatiion and rotation...

resource documents are vague and like many values in maya the ranges are hard to anticipate from one to the next.

[joojaa]

One way is to first fix the disortion in post then animate on that (track whatever) then instead of using this footage use the same routine in reverse on the cg elemnts and composite.

you could do the disortion in maya and many other porgs too but it requires allmost allways raytracing so for simple (raytracing wastes huge amounts of memory) renderings doing it in post is more effective.

but to do thsi realy vcreate a lens in front of the camera that acts as the disortion would (knowing optics is quite essential here)

[latrommi_suineg]

okay, youre missing the point... im not looking for post production tricks to distort my images... i can leave that to my compositor, he (they) are all quite good at there jobs...

what im asking for is explanations beyond the included maya documentation that can offer me a means to simulate a real world camera inside maya... the variables are too many and there is no clear min max ranges to easily formulate a working solution... i have an idea of the functions of each attribute and how they effect each other... but am always surprised at the difference that a 3d tracking software produces when i import the camera into my scene... and im wondering why in softimage the persp cameras were so much more sensitive to each incremental change and tweak, and why the maya cameras look like orthographics.

please.

[joojaa]

Maya doesent do disortion caused by roundness of lenses unless you place a lens in front of the camera! And since its a geometry issue it is not explained. Other then that they work exactly like a relaworld camera.

you probably want to adjust the fov, theres a good tutorial at the CEC page (dont remeber the url=, if you want more detailed info.

[latrommi_suineg]

i dont mean to be critical of your assistance but your replies are often hasty and seem thoughtless...

although im sure you have a wealth of experience to offer... im not finding it in the many comments that you leave accross the board... you seem to be more interested in answering questions than listening to them. here is mine:

and yes maya does do camera distortions like real world cameras... its a question of properly combining angle of view, and focal length... however, as i have said... its hard to guess at min max ranges because there are no global attributes in maya, what needs increments of 0.1 to 1 for one attribute might need 100- to 1000 ranges for another (for example light intensity... and photons)...

what i was hoping to do was dodge taking a camera training program for some simple guides in maya which i might lean on to push my images towards greater realism...

joojaa, if you have something worth saying... take the time to be clear and be exact... CEC= what?... i could look on the internet for days and come up with nothing

[joojaa]

Turned into a rant that might or might not help you! Sorry about that, ill go into this a bit deeper on monday since im leaving now for the weekend.

Ahh CEC url is http://cade.scope.edu/ and they have some brief discussions on this (pertty good). And yes you wouldt find this on the net because they have a flair for using no text in navigation so searcheingenes get very hung up on this.

But to be honest from your first post i tought you had problems with the lens disortion (since its a very common issue during tracking). And it could have well ben this you were talking about

Well yes im not the definitive pro on the subject on camera filmback and such (im more of aprocedural shaders guy and a particle dynamics worker). And so this explanaition would require a ton of pictures. But its not allways you have huge amonts of time to dedicate on exact workings of things. you can push people on the way but you cant do their jobs. By the way i swiched tactics from trying to explain wery troughly to brief answers about a year ago. Since nobody understod anyway (except for the few who get the drift with brief posts anyway). the CEC site site explains tha camera values pretty well, and the rest you get from the maya manual especialy under the maya live documentation, so i will not go in there so much

But about units here goes some rant: To be honest the camera values in maya are real world units (as long as your pref units is in scale, oh it doesent realty matter sice cameras work the same way independent of scale) thats why they are so hard to increment. the filmback and focal length is in mm etc. That is also the reason why your other values get so high. This is very common in realworld physics. They all make sanse if you get the units they are talking about the filmback is in mm as is the focal leght the angle of view is in dergees so too big values wouldnt make any sense. While photonmapping is number of pohotons to be shot and so 0.1 would be quite out of the question (in realworld you get bombarder by millions of photons all the time, not to mention those that dont hit the retina).

The only thing here that is realy (kindoff) unitless is the light intensity, wich is reported in screenspace maxima (white with intensity of 1 would be 255 255 255 in rgb), since we are not using floating point output files wich would be more inteligent. Altough there are many things that affect this such as dropoff and surface color surface fallof etc. So a truly white lambert surface with white light of 1 1 1 rgb with intensity of 1 with no fallof then the points in the surface that face the light would be 255 255 255 white ina 8 bit image.

By the way what has light intensity to do with camera settings by the way?

Hope this get you on your way, no offense taken by the way

[sumo-kun]

The maya camera has only a few parameters that you can adjust. You basically can only adjust the size of the filmback (as an aspect ratio, the actual numbers really do not matter. YOu can also adjust the aspect of how this film back is rendered to a pixel image (rect pixels vs square, fully adjustable)) and the distance from the "lens". The important thing to understand here is that the lens is actually a pinhole, with no further parameters. This is also the same for softimage and most other scanline renderers (as far as I know) Any further distortion is done via a post process (again, perhaps there are exceptions, I do not know how renderman deals with camera shaders, I only know they exist)) With mental ray, you can use a lens shader coupled with raytracing to simulate a physical lens. You can also fake this in maya using raytraced geometry in front of the lens and using either refraction through a "lens" or reflections in a spherical object. Either way, this is difficult to control, as there is no way to see accurate realtime feedback. I have also used a flare deformer parented to the camera affecting every object in the scene with good results. Then you get the feedback, but it can be tricky with large scenes.
For you purposes of matching live footage to CG, render you images larger than your output size (1.5 to 2 times) and have your compositors perform a distortion to match to the live plate. For most productions that will be doing this, it can be very helpful to shoot reference grid to asses the lens distortion. Also, make sure that when the telecine it, the frames that you work with are scanned image center = film center, a small unknown pan-n-scan can really mess you up.
Good luck, I hope they did not use a zoom.
-MD