Loglet SIFT for Part Description in Deformable Part Models: Application to Face Alignment

Qiang Zhang and Abhir Bhalerao

Abstract

We focus on a novel loglet-SIFT descriptor for the parts representation in the Deformable Part Models (DPM). We manipulate the feature scales in the Fourier domain and decompose the image into multi-scale oriented gradient components for computing SIFT. The scale selection is controlled explicitly by tiling Log-wavelet functions (loglets) on the spectrum. Then oriented gradients are obtained by adding imaginary odd parts to the loglets, converting them into differential filters. Coherent feature scales and domain sizes are further generated by spectrum cropping. Our loglet gradient filters are shown to compare favourably against spatial differential operators, and have a straightforward and efficient implementation. We present experiments to validate the performance of the loglet-SIFT descriptor which show it to improve the DPM using a supervised descent method by a significant margin.

Session

Posters 1

Files

Extended Abstract (PDF, 1M)
Paper (PDF, 8M)
Supplemental Materials (PDF, 613K)

DOI

10.5244/C.30.31
https://dx.doi.org/10.5244/C.30.31

Citation

Bibtex

        @inproceedings{BMVC2016_31,
        	title={Loglet SIFT for Part Description in Deformable Part Models: Application to Face Alignment},
        	author={Qiang Zhang and Abhir Bhalerao},
        	year={2016},
        	month={September},
        	pages={31.1-31.12},
        	articleno={31},
        	numpages={12},
        	booktitle={Proceedings of the British Machine Vision Conference (BMVC)},
        	publisher={BMVA Press},
        	editor={Richard C. Wilson, Edwin R. Hancock and William A. P. Smith},
        	doi={10.5244/C.30.31},
        	isbn={1-901725-59-6},
        	url={https://dx.doi.org/10.5244/C.30.31}
        }