n the framework of the PRIME research project, the following Raw Data Workflow has been developed by Fraunhofer IIS.

1. Recording

Two microHDTV cameras have been installed on a side-by-side rig as a stereo setup. Both the cameras deliver crude raw data from the sensor to their HD-SDI Single Link outputs. The frame rate is 24fps per camera at a resolution of 1920*1080. The cameras are synchronized over the output of the Fieldrecorder Megacine.

2. Storage

Each camera is connected with a HD-SDI input (Single Link) of the Megacine. The Megacine records the raw data without any processing. Since the data is only available in Bayer format, the Megacine’s preview screen can only show a gray scale preview of one of the two camera images. For zooming in, a HD screen (with detail enlargement) would make sense.

3. PC

Through a IT port (Fibre Channel or FireWire) the Megacine is connected to a PC after the taping. The computer recognizes the Megacine as its own drive. By means of m-transfer software raw data is converted into dng-format by the Megacince and stored on the harddrive. (Dng format is supported by some post production tools.)

4. Algorithm

Previous to the post production the purely technical processing of the image data takes place on a PC. This is done on the dng raw data. It includes the color balance of the different cameras (marginal differences between the cameras) and the color calibration of both cameras towards one goal (marginal color difference to the goal), respectively. Result of the color balance is a 3x3 matrix which is applied to the raw data. Additionally the image data are geometrically corrected (intra- as well as inter-camera). During this equalization a 3x3 matrix plus 4 parameters (2 for radial and 2 for tangential distortion) are calculated, which are applied to the raw data. As the last algorithmic point, the initial depth estimation on the basis of the external calibration data takes place if necessary for the output medium. The result of this is a depth map for every r-1-image pair.

5. Postproduktion

The post production takes place on the basis of dpx data. For this the dng data have to be de-bayered and converted into dpx format. Editing and Color Grading are done by means of a 3D-compatible post production software. This is when the cameras’ convergence is ultimately set (it is established which objects are on the camera screen level, which ones in front of/behind it, respectively). For the evaluation of the 3D effect it has to be possible to watch in 3D in the editing suite.

6. Authoring

By means of the 3D-compatible software easyDCP the dpx data is packed into a JPEG2000-coded 3D DCP.

7. Presentation

The 3D DCP can be played on 3D-compatible D-Cinema Servers and the different 3D systems

The outline for the specification of a raw data format for HD-SDI Single Link connections can be found here.