Quality Assessment of Stereoscopic Images Using Deep Learning

Deljit, Nivedya and Channappayya, Sumohana (2019) Quality Assessment of Stereoscopic Images Using Deep Learning. Masters thesis, Indian institute of technology Hyderabad.

Text
Thesis_Mtech_EE_5463.pdf - Submitted Version
Restricted to Repository staff only until June 2021.
Download (1MB) | Request a copy

Abstract

For a stereoscopic image, quality is mainly contributed by left view, right view and depth/disparity map. However due to the unavailability of ground truth disparity map, most stereo quality assessment algorithms make use of estimated disparity map which makes the algorithm error prone. We propose a quality assessment method that makes use of disparity information without explicitly using a disparity map. An intermediate image is formed from the left and right views of a stereo pair using an Unsupervised Image to Image Translation (UNIT) framework. Through this intermediate image which we call the latent image, we can move from left view to right view and vice versa. This makes us believe that this intermediate image has some disparity/depth information and hence use it for quality estimation. Experimental results prove the same.

[error in script]

IITH Creators: