Computer-assisted endoscopy is a rapidly developing eld of study. Many image anal- ysis algorithms exist, achieving very high rates of eciency at processing single endoscopic images. However, most of them were never tested in processing real-life endoscopic videos. In the article such tests of 16 endoscopy image analysis algorithms are presented and dis- cussed. Tests were performed on two real-life endoscopic videos of a human large bowel. Each algorithm was trained and tested twice: for recognizing cancer/normal tissue and for blurry/sharp frames. Results of those tests showed that there exists only a small correlation between algorithms' eciency scores (from single images tests) and their performance at pro- cessing real-life endoscopic videos. This brought to conclusion that eciency tests on single images are not sucient for estimating algorithms' quality in real-life videos processing.
Authors
Additional information
- Category
- Aktywność konferencyjna
- Type
- publikacja w wydawnictwie zbiorowym recenzowanym (także w materiałach konferencyjnych)
- Language
- angielski
- Publication year
- 2013
