With the advent of the twenty first century, stereoscopic systems have found a widespread use in the engineering industry. Several biomechanical analyses utilize this concept for efficient information extraction. Some examples of its applications are gait analysis, hand shape recognition, facial surface recognition etc.

The primary goal of this thesis was to optimize the existing stereoscopic system, to increase accuracy and precision of the depth information extracted. The process included redesign of the existing equipment set-up, automation of image acquisition unit and modification of conventionally used correspondence test to achieve higher accuracy. The acquired data was used to estimate depth of the object used for the study.

It was found that the correspondence information for a pair of adjacent cameras had high accuracy. In addition, the plots of the correspondences exhibited similarity in the trend. An attempt was made to use this information for predicting the values for regions on the curves having inconsistencies. Depth estimation using triangulation was performed on the correspondences found for adjacent pairs. It was found that the row was reconstructed as anticipated using the algorithm. With further development of the algorithm and successful implementation of knowledge propagation, this system can demonstrate efficient shape recovery.