Available Knowhow
Quadratic Chi histogram family increases accuracy of feature matching
Categories |
Imaging / Computer Graphics, Computer Vision, Data Mining |
Development Stage |
Completed research. Proof of concept |
Knowhow |
Available Knowhow |
Highlights
-
Quadratic Chi (QC) histogram distance family is a family of algorithms that improves the accuracy of comparing and matching features between images.
-
Accelerates database retrieval
-
QC distances can be employed in computer vision applications such as tracking
Our Innovation
Novel family of algorithms that improves the measurement of distances between histogram bins.
In the four color histograms, each histogram has four colors: red, blue, purple, and yellow. The Quadratic-Form (QF), the Earth Mover Distance (EMD) and the L1 norm do not reduce the effect of large bins. Members of the Quadratic-Chi histogram distance family, QCN and QCS consider (a) to be most similar, (b) the second and (c) the least similar as they take into account cross-bin relationships and reduce the effect of large bins, using an appropriate normalization.
Additional information can be found at http://www.cs.huji.ac.il/˜ofirpele/QC/
Key Features
-
Achieves both robustness and distinctiveness
-
Reduces the effect of differences caused by bins with large values
-
Takes cross-bin relationships into account
-
Quadratic Chi members outperform state of the art distances for image retrieval using the Scale Invariant Feature Transform (SIFT) and color image descriptors as well as shape classification using Shape Context (SC) and Inner Distance Shape Context (IDSC) .
-
Short running time
The Opportunity
-
Possible applications in computer vision, data mining, medical applications, online mapping application