Sketching Method Based on Earth Mover's Distance for Image Contour Matching

摘要

Finding similar images to a given query image can be computed by different distance measures. One of general distance measures is the Earth Mover's Distance (EMD). Although, EMD has proven its ability to retrieve similar images in >95% true, high execution time is its major drawback. Therefore, previous algorithms of EMD could not run efficiently when performing retrievals from large databases. A contour-matching algorithm has been presented that quickly estimate the minimum weight matching using an embedding of the EMD into L,. This low-distortion algorithm somehow solves the time problem by sacrificing the performance due to generating heavily tailed image feature vector.