结合视觉感知与LBP傅里叶直方图的图像检索-模式识别.PDFVIP

23 3 V ol . 2 3 N o. 3 2011 3 Jo ur nal of Co mputerA ided Design Computer G raphics M ar . 2011 LBP 黄传波, 金 忠 ( 2 10094) ( huangjunf engcq@ 126. com) , , ( L BP) . , , , I tti , 50 ; L BP , ; ( L PP ) , . , , . ; ; ; T P39 1. 41 Image Retrieval Based on Combination of Visual Perception and Local Binary Pattern Histogram Fourier H uang Chu anbo and Jin Zhong ( College of Comp uter Science and Technology , N anj ing University of Science and Technology , N anj ing 2 10094) Abstract In sp ired by the f eat ure ext raction during t he bio logic v isual inf orm ation pro cessing, an algorit hm b ased on a com binat ion of v isu al p ercept ion and local binary p att ern histo gram F ourier ( LBP H F) is proposed t o ef fect ively ext ract t he f eat ures t o im pr ove the perfo rmance of imag e ret riev al. First ly, pr incip al component map is u sed as the pr im ar y visual f eat ure of int en sit y . T he inf orm at ion of sh ape and edge is f ur ther f used t o impr ove t he It tis model. T herefor e, w e can obtain 50 f eat ure m aps. Secondly, t he LBPH F o f each feat ure m ap s is com put ed and concat enat ed t o g et an enh anced image feature descriptor vector . T hirdly, t he locality preserv ing projection s ( L PP) is ut iliz ed fo r dim en sionality redu ct ion, and the f inal low dimensional f eat ure is u sed f or image retr ieval. T he exp erim ent al r esult s show t h at our method h as the discriminat ion pow er against color, tex ture and sh ap e features and h as goo d ret rieval p erfo rmance. Key words im age r et rieval; visu al att ent ion ; visual f eat ur e map ; local binar y pat t ern hist ogr am Four