distance transforms of sampled functions.pdfVIP

Distance Transforms of Sampled Functions Pedro F. Felzenszwalb Daniel P. Huttenlocher The University of Chicago Cornell University pff@cs.uchicago.edu dph@cs.cornell.edu Keywords: Euclidean distance transform, Minimum convolution, Dynamic Programming. Abstract This paper provides linear-time algorithms for solving a class of minimization problems in- volving a cost function with both local and spatial terms. These problems can be viewed as a generalization of classical distance transforms of binary images, where the binary image is replaced by an arbitrary sampled function. Alternatively they can be viewed in terms of the minimum convolution of two functions, which is an important operation in grayscale mor- phology. A useful consequence of our techniques is a simple, fast method for computing the Euclidean distance transform of a binary image. The methods are also applicable to Viterbi decoding, belief propagation and optimal control. 1 Introduction Distance transforms are an important tool in computer vision, image processing and pattern recognition. A distance transform of a binary image specifies the distance from each pixel to the nearest non-zero pixel. Distance transforms play a central role in the comparison of binary images, particularly for images resulting from local feature detection techniques such as edge or corner detection. For example, both the Chamfer [5] and Hausdorff [12] 1 matching approaches make use of distance transforms in comparing binary images. Distance transforms are also used to compute the medial axis of digital shapes [3]. In this paper we consider a generalization of distance transforms to arbitrary functions on a grid