E. Spyrou, and D.K. Iakovidis, “Video-Based Measurements for Wireless Capsule Endoscope Tracking,” Measurement Science and Technology, vol. 25, pp. 015002 (14pp), IOP Publishing, 2014

The paper is available here. 

How to construct the dataset used in this study:

·     The endoscopic data used in this study originate from the Given Imaging Atlas, which is a publicly available database of endoscopic images and videos. Access can be obtained after a simple registration process. To download the data visit the following website:  http://capsuleendoscopy.org

·     The data used in this study can be obtained using the following uniform resource locator (URL) template, with the previously mentioned website as root: /_layouts/Pages/CEPages/FileLoader.aspx?fileId=###&attach=true

·     Placeholder ### represents the file identification numbers which are available through this text file: fileId.csv

·     Install ffmpeg: http://www.ffmpeg.org/

·     Once a video clip, e.g. “inputfile.avi” (-i option) is downloaded in a folder, in order to extract frames at a rate of 1 fps (-r option) with filenames “frame-001.png”, “frame-002.png...” (-r option) execute in that folder the following: $ ffmpeg -i inputfile.avi -r 1 -f image2 frame-%3d.png

·     Execute in the same folder the following Matlab script that can process those downloaded videos into the form that was used in this work i.e. the rotations and scalings so that the dataset was exactly the same to what they used: create_data_set.m. The naming of the processed frames will be of the format e.g. “frame-001_sc_0_2_rot_10” i.e. frame-001.png rotated by 10 degrees and scaled by 0.2