Edge detection in discretized range images

In this paper we show a method for edge detection in discretized range images. Heavy quantization results in low number of layers present in range images. These values suffer from systematic error and must not be treated as random additive noise. In our method we utilize morphologic operations for noise noise reduction and skeleton extraction for each layer. Based on the extracted features such as layer skeletons we present an algorithm to identify and classify edges in range images. We present edge detection results of simulated and captured real range images.