One application of LLE in CBIR
Lan Zhang · Journal of Science of Teachers' College and University · 2009
In content based image retrieval,the dimension of an feature vector will increase seriously when more details of the image are given.The spratial complexity of high eigenvector's storage and operation of computing their similarity will be much higher.An image is decomposed by a color homogeneity predicate to get the feature of color,and LLE is applied to the dimension reduction to reduce both the spatial and operation complexity.Experimental results shown that LLE united with this way of abstract the feature of an image have better results.