Rényi Divergence As Contrast Metric For Object Detection In Noisy Images
Résumé
A binary detection task on the presence or absenceof an object over a background in noisy images is studiedbased on a contrast metric taken as the R´enyi divergence. Theanalysis extends, and contains as a special case, previous worksrealized for the same detection task on the comparison of theKullback-Leibler divergence and the Bhattacharryya divergence.The extension deals with the influence of the R´enyi parameter,the observation scale, the influence of the bit range and the typeof noise tested with new noise families of practical interest forimaging.