Depth map reconstruction from stereo image using sum of absolute differences

Sabtuhe, Effy Shazrin (2021) Depth map reconstruction from stereo image using sum of absolute differences. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

	Text (24 Pages) Depth map reconstruction from stereo image using sum of absolute differences.pdf - Submitted Version Download (1MB)
	Text (Full text) Depth map reconstruction from stereo image using sum of absolute differences.pdf - Submitted Version Restricted to Repository staff only Download (2MB)

Abstract

This project provides a method for solving the issue of correspondence when matching the stereo image using the algorithm Sum of Absolute Differences (SAD). MATLAB software provides the tool. The mapping of disparities is produced through the block matching algorithm Sum of Absolute Differences (SAD). The SAD conventional is usually scanned in entire stereo images to figure out the discrepancy in lines between the captured images on the left and right, and then obtains the equivalent map of disparity, which can result in a high elapse time. There are four basic steps in a stereo vision method for imaging the reconstruction. There are typically four phases of this The distortion of the captured images from the camera lens is extracted first in the un distortion step. The next step is to adjust the distance and the elevation angle between the two camera images to determine the focal length and the epipolar axis, as presumed beyond. The comparisons inter the left and the right image will be calculated in the correspondence stage and used to measure the map of disparities. This method is often known as being interoperable. This project was able to develop stereo matching algorithm using sum of absolute differences to do the depth map reconstruction from the proposed algorithm and a simple stereo algorithm compute results comparable to current state-of-the-art on Middlebury benchmark. The method of this project is using a graphical user interface that developed from MATLAB environment that compute the disparity mapping by using sum of absolute differences. This project has answered all my objectives and the results I got to achieve the set benchmark. In various domains, such as tracking, robot navigation, etc., these algorithms can be better used. There is no better or indispensable depth cue than another depth cue. A cue has its own perks and disadvantages.

Actions (login required)

View Item