Novel hierarchical image retrieval using geometric invariants and normalized histogram
Liu Ji-guang, Jinlong Zheng · Journal of Lanzhou University · 2010
Hierarchical image segmentation was proposed to partition the retrieved images into equal blocks with different sizes at difficult levels.Then,the similar metrics of these sub-blocks to query image were evaluated to retrieve those sub-blocks with contents in query images.Meanwhile,information about scales and locations of query objects in retrieved images could also be retrieved.The hierarchical block image retrieval schemes using geometric invariants and normalized histograms were tested by experiments via a database with 500 images.The retrieval accuracy with geometric invariants as invariant features could achieve 78%for the optimal similar metrics threshold.Furthermore,the schemes could also work with different size images.