Multistereo system design

Depth estimation using stereoscopy can be done by comparing the relative positions of the same (point) object displayed in the two different images captured by two cameras. Quantization errors that follow a square law with respect to distance are one of the errors that arise during depth estimation. The solution to limit this error is to use a multi camera arrangement. We deal with the physical aspect of arrangement of a multi camera system to reduce quantization error of the camera system, and to make the quantization error more uniform with respect to distance (over a certain range), using multiple baseline stereo. We also propose a method to calculate the minimum number of cameras required for a given camera resolution to measure a range of depth with a bounded error performance. The most important contribution is our investigation of the camera placement problem: given a set of pairwise camera distances, does a solution even exist? It turns out that a solution does not exist in general. Further, we deal with considerations pertaining to the arrangement of the cameras to find the direction as well as the depth of the object with respect to the camera system.