Abstract

The hexagonal structure is considered to be preferable to the standard rectangular structure typically used for images in terms of the improved accuracy and efficiency that can be achieved for a number of image processing tasks. However, due to the lack of commercially available hexagonal image sensors, hexagonally structured images are generated by resampling from standard rectangular images and hence many resampling techniques exist for this purpose. Here we consider four such resampling techniques and apply recently developed scalable operators to them for the purpose of feature extraction. Each hexagonally structure image is evaluated with respect to feature extraction performance and we provide conclusions on the most accurate resampling technique currently available.