Abstract: Simplification has always been a commonly-used generalization operator, and a number of simplification algorithms are currently available. It is natural to further pay more attention to the performance evaluation of these algorithms. With regard to this, there are some representative evaluation indicators which are proposed with the consideration of the differences of the location and/or shape of spatial features. Indeed, these individual indicators only consider some aspect of the distortions caused by generalization, so that it is difficult to comprehensively reflect the performance of simplification algorithms. To overcome such problem, this paper takes line feature as an example and develops a new evaluation indicator (i.e. information content) at three levels (i.e. element, neighborhood and holistic levels) from the view of information transmission. The process of performance evaluation mainly involves the calculation and the comparison of information content before and after line simplification. As for the former, the quantitative computational methods of information content at the three levels are presented. As for the latter, the difference of information content before and after line simplification is computed and utilized to measure performance of line simplification algorithms. The difference degree of information content at element level reflects the ability of line simplification algorithms to select key points; the difference degree of information content at neighborhood level reflects the ability of line simplification algorithms to maintain bends; the difference degree of information content at holistic level reflects the ability of line simplification algorithm to maintain the trend of line features. Finally, a river network dataset is used to test the performance evaluations of four common-used line simplification algorithms according to the proposed indicator of information content. It is proven that this new indicator is very rational to evaluate the performance of these four line simplification algorithms. At the meantime, a comparative test is made to show the advantages of the new evaluation indicator.

Key words: performance evaluation, hierarchy, information content, line simplification, map generalization