. . "Which stereo algorithm is better? The one which is very dense but often erroneous or rather one which is very accurate but sparse? It depends on the application. In general, we can only say that the algorithm is better than the other if it is more accurate and denser. In literature, there exist several methods to evaluate quality of dense stereo matching algorithms. Their bottleneck is in tested algorithm parameter setting, which is assumed to be fixed. Such evaluation results are typically very different for various parameter setting in the sense they somehow change the tradeoff between accuracy and density. Therefore, we developed a new method for testing stereo algorithm based on the ROC-like analysis. We introduce ROC curves for stereo algorithms and define new numerical characteristics, which evaluate the algorithm itself, not a pair (algorithm, parameter setting) as it is in literature. Comparing ROC-curves of all tested algorithms we obtain the Feasibility Boundary."@cs . . . . "ROC analysis; computer vision; dense stereo; performance evaluation"@en . . "P(1ET101210406), R" . "ROC Based Evaluation of Stereo Algorithms" . . "[D8D49BE3DF45]" . "RIV/68407700:21230/07:03134573" . "ROC Based Evaluation of Stereo Algorithms" . "ROC Based Evaluation of Stereo Algorithms"@en . "ROC Based Evaluation of Stereo Algorithms"@en . "RIV/68407700:21230/07:03134573!RIV09-MSM-21230___" . . . . . "ROC Based Evaluation of Stereo Algorithms"@cs . . . . . "3"^^ . "Which stereo algorithm is better? The one which is very dense but often erroneous or rather one which is very accurate but sparse? It depends on the application. In general, we can only say that the algorithm is better than the other if it is more accurate and denser. In literature, there exist several methods to evaluate quality of dense stereo matching algorithms. Their bottleneck is in tested algorithm parameter setting, which is assumed to be fixed. Such evaluation results are typically very different for various parameter setting in the sense they somehow change the tradeoff between accuracy and density. Therefore, we developed a new method for testing stereo algorithm based on the ROC-like analysis. We introduce ROC curves for stereo algorithms and define new numerical characteristics, which evaluate the algorithm itself, not a pair (algorithm, parameter setting) as it is in literature. Comparing ROC-curves of all tested algorithms we obtain the Feasibility Boundary." . "\u010Cech, Jan" . "21230" . "Kostliv\u00E1, Jana" . "3"^^ . . . "Which stereo algorithm is better? The one which is very dense but often erroneous or rather one which is very accurate but sparse? It depends on the application. In general, we can only say that the algorithm is better than the other if it is more accurate and denser. In literature, there exist several methods to evaluate quality of dense stereo matching algorithms. Their bottleneck is in tested algorithm parameter setting, which is assumed to be fixed. Such evaluation results are typically very different for various parameter setting in the sense they somehow change the tradeoff between accuracy and density. Therefore, we developed a new method for testing stereo algorithm based on the ROC-like analysis. We introduce ROC curves for stereo algorithms and define new numerical characteristics, which evaluate the algorithm itself, not a pair (algorithm, parameter setting) as it is in literature. Comparing ROC-curves of all tested algorithms we obtain the Feasibility Boundary."@en . . . "448093" . "ROC Based Evaluation of Stereo Algorithms"@cs . "\u0160\u00E1ra, Radim" .