Problems in distortion corrected texture classification and the impact of scale and interpolation

Abstract

In the field of computer aided celiac disease diagnosis, wide-angle endoscopy lenses are employed which introduce significant barrel type distortions. Although the images can be rectified using distortion correction methods, computer based diagnosis suffers from missing information in highly distorted image regions. First, we investigate the impact of simple and advanced interpolation techniques on the classification rates. Furthermore we explore the effect of considering different image resolutions. Whereas in previous studies distortion correction in most cases turned out to be disadvantageous, we show that for certain setups distortion correction definitely is advantageous. © 2013 Springer-Verlag.