Abstract:
In this paper, we discuss the disparity estimation problem which is the most important
challenge to generate an intermediate image for view synthesis. The main problem in
estimating the disparity is to find the correspondence among the input images. The
quality of the synthesized novel view is dependent on the accuracy of disparity
estimation. Once we find the exact disparities among the input images, a simple low
complexity interpolation is carried out to generate the new view. We offer a view of a
comparative study between different state-of-the-art disparity estimation algorithms
showing the advantages, disadvantages, and the peak signal to noise ratio (PSNR) of the
reconstructed view using each one.