Thumbnail
Access Restriction
Open

Author Xie, Yusheng ♦ Chen, Zhuoyuan ♦ Agrawal, Ankit ♦ Liao, Wei-Keng ♦ Choudhary, Alok
Source CiteSeerX
Content type Text
File Format PDF
Subject Domain (in DDC) Computer science, information & general works ♦ Data processing & computer science
Subject Keyword Certain Non-rigid Transformation ♦ Novel Idea ♦ Individual Image ♦ Numerical Vector ♦ Image Index ♦ Diversified Search Result ♦ Impractical Amount ♦ Efficient Algorithm ♦ Local Point Feature ♦ Tree-like Data Structure ♦ Unnecessary Nrdc Computation ♦ Existing Image ♦ Pt Structure ♦ Phylogenetic Tree ♦ Different Expression ♦ Scale Invariant Feature Transformation ♦ Non-rigid Dense Correspondence ♦ Large-scale Image Indexing ♦ Real World Application ♦ Fast Search ♦ Social Hashtags
Abstract Most existing image indexing techniques rely on Scale Invariant Feature Transformation (SIFT) for extracting local point features. Applied to individual image, SIFT extracts hundreds of numerical vectors. The vectors are quantized and stored in tree-like data structures for fast search. SIFTbased indexing can exhibit weakness under certain non-rigid transformations, which are common among real world applications. For example, SIFT often cannot recognize a face as the same with different expressions (e.g. giggling vs. crying). Non-Rigid Dense Correspondence (NRDC) addresses such drawbacks of SIFT. However, directly using NRDC incurs an impractical amount of computation in large-scale image indexing. We present a novel idea here that uses social hashtags to organize the images into a phylogenetic tree (PT). We provide an efficient algorithm to build/search the PT, and show that using PT structure can effectively avoid unnecessary NRDC computation. The resulting image index provides more accurate and diversified search results 1.
Educational Role Student ♦ Teacher
Age Range above 22 year
Educational Use Research
Education Level UG and PG ♦ Career/Technical Study