Image categorization based on EMD-CkNN multi-instance learning algorithm
HE Jing-fang · Journal of Optoelectronics·laser · 2010
Focusing on the problem of natural image scene classification and according to the idea of Citation-kNN algorithm,a new image categorization method is proposed based on multi-instance learning.This method regards the whole image as a multi-instance bag and the segmented regions as instances in this bag.When measuring the similarity between different bags,an improved earth mover distance instead of the minimum Hausdorff distance used in the Citation-kNN multi-instance learning algorithm is used for image categorization.Experimental results on the images in COREL dataset have shown that the performance of our algorithm is better than that of the original Citation-kNN algorithm.