EditorialPerformance characteristics of vision algorithms

For at least 10 years computer vision has been confronted with papers and discussions on the scientific value of its results and the difficulties in transferring the results to practical systems. A change of awareness has happened: More than 10 years ago, at the Computer Vision Workshop 1985, two controversial papers with different viewpoints, agreed on the lack of theoretical research [3, 7], which should go along with the development of vision procedures: experimental proofs are not enough. Five years ago, the dialogue on ‘Ignorance, Myopia, and Naiveté in Computer Vision Systems’ initiated by R. Jain and T. Binford [4] and the responses documented the necessity of evaluating theoretical findings, vision procedures algorithms etc. by using empirical data in order to increase the number of real world applications of computer vision research. When observing the increasing number of papers which propose new solutions to classical problems, especially using increasingly more demanding theoretical tools, it seems to become clear that empirical testing of vision algorithms is necessary to allow a clear comparison of the proposed methods by the users of such algorithms. Together with the underlying theories a clear performance characterization of algorithms is necessary. This special issue is motivated by the belief that the lack of performance characterization of vision algorithms is responsible for the hesitation of industry to use computer vision as one of its tools. Reasons for this situation are manifold: the lack of commonly accepted criteria for evaluation, the lack of a methodology for testing, the lack of translating the experience in testing of other engineering areas to computer vision and possibly also the non-acceptance of empirical or theoretical comparisons of vision algorithms, including their replication, as original research.