SEGMENTATION OF OCT SKIN IMAGES BY CLASSIFICATION OF SPECKLE STATISTICAL PARAMETERS

This paper deals with segmentation of dermatological OCT images. Classic image segmentation techniques fail to produce accurate results due to the wide presence of speckle. We propose using speckle as source of information in the segmentation process. Different statistical models are analyzed in terms of their ability to differentiate skin layers. The local speckle parameters are used as a features-vector to classify, in a supervised way, different regions. Experimental results are presented using a corpus of twenty three real images delineated by experts. These confirm the potential of the method to generate useful data for robust segmentation.