Descriptions:

The Waterloo-IVC 3D Depth Database is created from 6 pristine texture images (Bark, Brick, Flowers, Food, Grass, and Water). All images were collected from the VisTex Database at MIT Media Laboratory. Each texture image was used to generate different stereopairs with 3 depth polarizations (inner, outer, and flat) and 6 depth levels.

Each pristine stereopair (inner, outer, and flat) was altered by three types of distortions: additive white Gaussian noise contamination, Gaussian blur, and JPEG compression. Each distortion type had four distortion levels, where the distortion control parameters were decided to ensure a good perceptual separation. The distortions were simulated either symmetrically or asymmetrically. There are 72 pristine stereoscopic images and 1728 distorted stereoscopic images (864 symmetrical and 864 asymmetrical distortions) in the database. In terms of the depth polarity, there are 684 inner stereopairs, 684 outer stereopairs, and 432 flat stereopairs.

Publications:

• Jiheng Wang, Shiqi Wang, Kede Ma and Zhou Wang, "Perceptual depth quality in distorted stereoscopic images", IEEE Transactions on Image Processing, vol. 26, no. 3, pp. 1202-1215, Mar. 2017. [pdf]
• Jiheng Wang, Shiqi Wang and Zhou Wang, "Depth perception of distorted stereoscopic images", IEEE International Workshop on Multimedia Signal Processing (Top 10% award), Xiamen, China, Oct. 19-21, 2015. [pdf][slides]

Download Now: Database [1.17GB] and Matlab Code [62.1KB]

Password: uwivc3D2017