Stereo Image and Depth Map Generation for Images with Different Views and Resolutions

Abstract

There are many practical applications where stereo images and depth maps may need to be generated using cameras with different resolutions and view angles. One challenge is that the parallax in images may create issues in image alignment. Another challenge is on how to handle the different resolutions. This paper presents some preliminary results in enhancing the stereo images and depth maps using the uncalibrated images with different resolution and view angles. Our goal is to investigate whether or not we can combine the images with different resolutions and view angles to generate high quality stereo images and depth maps. We first enhance the low resolution image (assuming it is the left image without loss of generality) with help from the high resolution image (assuming it is the right image). We then combine the improved left image and the original high resolution right image to create high quality stereo and depth maps. Actual images from the Middlebury database were used in our experiments. Preliminary results demonstrated that the proposed approach works to certain extent. There is still room for performance improvement.