Stereoscopic Image Quality Assessment by Yong Ding & Guangming Sun

Author:Yong Ding & Guangming Sun
Language: eng
Format: epub
ISBN: 9789811577642
Publisher: Springer Singapore

(5.57)

where Nhist is the number of feature groups, and are the number of pixels in the ith feature group of the left and right views. Thus PUP is defined as:

(5.58)

In this method, the width of the image patch should be set reasonably to capture large disparities. That’s because images with highly variable disparity levels and rich details will be better analyzed by adopting a smaller patch width, while images with few disparity levels and details will be better analyzed by adopting a larger width. Hence, we need multiple sizes of patch widths for accurate modeling. Generally, three different patches are adopted in this model. The width of the largest patch corresponds to the disparity limit of the fusional area (Yeh and Silverstein 1990). The width of the smallest patch corresponds to the disparity limit of the comfort zone (Percival 1892). The width of the average patch corresponds the average value of the fusional limit and the comfort limit disparities. Specifically, for an image with a resolution of 1920 × 1080 pixels, the corresponding block widths of PUP-L, PUP-M, PUP-S are calculated to be 280, 168 and 56 pixels respectively. Finally, to predict the degree of visual discomfort of stereoscopic image, each of four kinds of features are extracted from the PUP-L, PUP-A and PUP-S maps respectively with the feature extraction method in Park et al. (2014). Therefore, considering three different PUP maps, the total number of features used in prediction is 12. Four kinds of features contain the average values of positive and negative PUP, and the upper and lower 5% of the PUP values, which are defined as follows:

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.