Hyperspectral Imaging: Techniques for Spectral Detection and Classification
Hyperspectral Imaging: Techniques for Spectral Detection and Classification
An introduction to ROC analysis
Pattern Recognition Letters - Special issue: ROC analysis in pattern recognition
Extending Diabetic Retinopathy Imaging from Color to Spectra
SCIA '09 Proceedings of the 16th Scandinavian Conference on Image Analysis
| Hi-index | 0.00 |
Conventional 3-channel color images have limited information and quality dependency on parametric conditions. Hence, spectral imaging and reproduction is desired in many color applications to record and reproduce the reflectance of objects. Likewise RGB images lack sufficient information to successfully analyze diabetic retinopathy. In this case, spectral imaging may be the alternative solution. In this article, we propose a new supervised technique to detect and classify the abnormal lesions in retinal spectral reflectance images affected by diabetes. The technique employs both stochastic and deterministic spectral similarity measures to match the desired reflectance pattern. At first, it classifies a pixel as normal or abnormal depending on the probabilistic behavior of training spectra. The final decision is made evaluating the geometric similarity. We assessed several multispectral object detection methods developed for other applications. They could not proof to be the solution. The results were interpreted using receiver operating characteristics (ROC) curves analysis.