Gradient operators for feature extraction and characterisation in range images

Abstract

In recent years range images have become prominent in computer vision applications as they provide an almost 3-D description of an otherwise 2-D scene and are suitable for computer vision tasks such as localisation and navigation. Feature extraction from range images has proven to be a complex problem; developing operators that can characterise features in a range image, such as step, crease, or smooth edges, is challenging, due to both the irregular spatial distribution of range image data and the nature of the features themselves. We present an adaptive design procedure for first order gradient operators that can automatically change shape to accommodate irregular data distribution; through appropriate analysis of the output responses, we show that the operators can also be specialised to characterise particular types of range image features. Hence the method is appropriate for direct use on range image data without re-sampling.