Tobias Reichl

Computer-Assisted and Robotic Endoscopy: Second International Workshop, CARE 2015 Held in Conjunction with MICCAI 2015 Munich, Germany, October 5, 2015 Revised Selected Papers

In a large experimental study, the impact of lossy image compression standards on CAD support systems based on texure classification is assessed using colonoscopic imagery as an example. Results clearly indicate that (1) it is important to compress both training and evaluation data involved in the classification process, (2) there is a big difference if initial data is precompressed or uncompressed, and (3) in the latter case significant improvements in terms of classification accuracy may be achieved, even and especially in case of high compression ratios. Moreover it is found that compression efficiency in terms of image quality metrics and/or human perception is not correlated with the impact compression has on texture classification accuracy.