Flinders Academic Commons (Flinders University) · 2005 · 15 citations · 0 references
Open access
ComputingComputer ScienceAlgorithmsChannel CodingCoding TheoryError-correcting CodesNearest Neighbor SearchSearching AlgorithmsDecoding AlgorithmsHadamard CodesNearest Neighbor Retrieval
This paper presents a fast codeword search algorithm that performs the equalaverage equal-variance equal-norm nearest neighbor search (EEENNS) in the ordered Hadamard transform (OHT) domain.By reordering the rows of Hadamard transform matrix, we can obtain the OHT with better energy packing efficiency, which is very important to the partial distance search (PDS) stage.Four elimination criteria based on three characteristic values, the first element, variance, and norm of the transformed vector, are introduced to reject a large number of unlikely codewords.Experimental results show that the proposed OHTEEENNS algorithm outperforms most of existing algorithms in the case of high dimension, especially for high-detail images.