Motion controlled dolly : video results

This is a description of the tests made for reproducing a 3D camera movement in stop motion animation. The tests are part of researches made to simulate a realistic camera movement in stop motion animation. They involve the CAGD-CGAO and Calhiste laboratories of the University of Valenciennes.

For all the tests, it all starts in 3Ds Max, making a very simple camera animation : a free camera makes a 1,35 meters travelling on its X axis in 250 images (25 s in PAL system) with an acceleration at the beginning and a deceleration at the end.

The two following figures show the animation in 3Ds Max viewports and the corresponding curve - its "S" shape indicate the acceleration at the beginning and the deceleration at the end.

anim_freecam

The free camera animation travelling (3Ds Max) : 1 square = 10 cm

anim_freecam_curve

The corresponding motion curve : the "S" shape indicate an acceleration at the beginning and a deceleration at the end. On the graph, X axis = time (in number of frames), Y axis = distance (in cm)

Once the animation is made, the camera's positions for each frame are exported in a text file with a home-made 3Ds Max script

mcsm_ui

User Interface of the script exporting the camera's positions to a text file

First Test : Full handmade stop motion travelling

Embed Embed this video on your site

Method : the X distance of the camera is written down on a paper for each frame. With a real camera (Olympus Camedia C3000), a photo is made at each position. The whole process is made by hand : positionning and shooting.

Result : Extremely shaky because 1) hand shooting photos add some random noise at each frame and breaks the continuity of the movement 2) because of the imprecision of the placement on X axis. A detail of the camera animation shows the cause of the second problem :

The red line shows the camera path, and each white point shows the position of the camera for each frame of the sequence. In this figure, one square equals to one centimeter : we can see that the first positions of the move are very close from each other. Distance between frame #0 and #1 is 0,06 mm, far below the human possible precision ! The use of a motion controlled camera is necesseray to obtain such a precision.

Second test : A motion controlled dolly

Embed Embed this video on your site

method : We've built a PC-controlled dolly, based on a step motor and a demultiplication gear system. Moving the motor of one step gives the dolly a less than 0,08 mm translation move (17.535 steps for 1,35 meters = 0,077 mm), allowing much more precision than in the previous test.

Result : Much less shaky because the second problem has been wiped out by the use of the PC controlled dolly, which is much more precise than hand positionning. But the result is still noisy because we've used the same camera than in the first test (Olympus Camedia C-3000) and it can't be controlled by the PC : all shoots are hand made, and each time the camera button is pressed, the camera moves a little eventually adding noise to the movement.

Third test : A full PC-controlled process

Embed Embed this video on your site

Method : Same than the previous test, but this time we used a Canon Eos 450D, controlled from the PC with DSLRpro. The whole sequence has been shoot in HDR mode (i.e 5 exposures for each frame) resulting in 1250 photos ! These ones are finally compiled in an HDR sequence with HDRShop and After Effects. Only the beginning of the movement is shown.

Result : This time, the noise has definitly been wiped out and the resulting HDR sequence gives a full 32bits floating point informations which is ideal for compositing process. The final result has a definition of 4272 x 2848 pixel - far better than the 1080 HD resolution. Such a high definition gives a lot of freedom for both compositing and broadcast.

Fourth test : Stop motion and integrated 3D objects

Embed Embed this video on your site

(3dsmax / Paricle Flow / Mental ray)

Method : Same than in the previous test, making some few more photos to obtain more informations on the scene (lightprobe, greyscale, colorscale, shadows properties, white balance).

Result : A ball explosion and a "Crealyse" word rendered from 3Ds max are composited in After Effects. The 3D rendering is well integrated in the stop motion sequence mainsly because during the whole workflow, pictures depth is 32bits floating point.