Difference between revisions of "Eyesis workflow"

Revision as of 17:05, 31 July 2010

Preface

The following page describes the Eyesis camera post-processing procedures.

Camera Output

Fig.1 Eyesis frames

• 3x 500GB HDDs, each containing *.mov files from 3x 3-sensor cameras (a single HDD is connected to a camera).
• Each frame in a *.mov file is a combination of frames from 3 sensors glued vertically.
• The cameras are numbered - 1,2,3. The sub images are [1.1, 1.2, 1.3], [2.1, 2.2, 2.3], [3.1, 3.2, 3.3]. 3.1 is the fisheye image.
• The frames in *.movs are indexed.
• Single frame format is JP4.
• Corresponding frames have the same time stamps generated by the 1st camera.
• The 1st camera frames are geotagged.

Post-Processing Setup

• HDDs are connected to a processing PC with SATA cables

Workflow

1. Building the Frames Index

• As the frames' index is stored in each corresponding *.mov, it is read and used for building the index of all the frames on 3x drives.

*.mov header's 'stco'-tag contains offsets for all the frames in the file.

2. Processing panorama

• Extracting and converting frames

* By index the 3 corresponding JP4 frames are copied from HDDs.
* JP4 -> DNG (or openEXR).
* Some Image Processing.

• Each frame is cut on 3 separate subframes - 9 images is the resulting array before stitching.
• Then goes stitching (of JPGs? each individual Eyesis has its constant gluepoints?) and the output is a panorama image.