A Biologically Inspired Approach for Fast Image Processing

As features within an image may be present at many scales, application of feature detectors at multiple scales can improve accuracy of the detected localisation and orientation. As the scale and size of a feature detector increases, so does the computational complexity of implementation across the image domain. To address this issue we present a novel integral image for hexagonal pixel based images and associated multi-scale operator implementation that significantly speeds up the feature detection process. We demonstrate that this framework enables significantly faster computation than the use of conventional spiral convolution, the use of a neighbourhood address look-up table on hexagonal images.