King's College London

Research portal

DCT-Inspired Feature Transform for Image Retrieval and Reconstruction

Research output: Contribution to journalArticlepeer-review

Standard

DCT-Inspired Feature Transform for Image Retrieval and Reconstruction. / Wang, Yunhe; Shi, Miaojing; You, Shan et al.

In: IEEE TRANSACTIONS ON IMAGE PROCESSING, Vol. 25, No. 9, 09.2016, p. 4406-4420.

Research output: Contribution to journalArticlepeer-review

Harvard

Wang, Y, Shi, M, You, S & Xu, C 2016, 'DCT-Inspired Feature Transform for Image Retrieval and Reconstruction', IEEE TRANSACTIONS ON IMAGE PROCESSING, vol. 25, no. 9, pp. 4406-4420. https://doi.org/10.1109/TIP.2016.2590323

APA

Wang, Y., Shi, M., You, S., & Xu, C. (2016). DCT-Inspired Feature Transform for Image Retrieval and Reconstruction. IEEE TRANSACTIONS ON IMAGE PROCESSING, 25(9), 4406-4420. https://doi.org/10.1109/TIP.2016.2590323

Vancouver

Wang Y, Shi M, You S, Xu C. DCT-Inspired Feature Transform for Image Retrieval and Reconstruction. IEEE TRANSACTIONS ON IMAGE PROCESSING. 2016 Sep;25(9):4406-4420. https://doi.org/10.1109/TIP.2016.2590323

Author

Wang, Yunhe ; Shi, Miaojing ; You, Shan et al. / DCT-Inspired Feature Transform for Image Retrieval and Reconstruction. In: IEEE TRANSACTIONS ON IMAGE PROCESSING. 2016 ; Vol. 25, No. 9. pp. 4406-4420.

Bibtex Download

@article{85a11e3fa17946dcaa46b6ed8774090b,
title = "DCT-Inspired Feature Transform for Image Retrieval and Reconstruction",
abstract = "Scale invariant feature transform (SIFT) is effective for representing images in computer vision tasks, as one of the most resistant feature descriptions to common image deformations. However, two issues should be addressed: first, feature description based on gradient accumulation is not compact and contains redundancies; second, multiple orientations are often extracted from one local region and therefore produce multiple descriptions, which is not good for memory efficiency. To resolve these two issues, this paper introduces a novel method to determine the dominant orientation for multiple-orientation cases, named discrete cosine transform (DCT) intrinsic orientation, and a new DCT inspired feature transform (DIFT). In each local region, it first computes a unique DCT intrinsic orientation via DCT matrix and rotates the region accordingly, and then describes the rotated region with partial DCT matrix coefficients to produce an optimized low-dimensional descriptor. We test the accuracy and robustness of DIFT on real image matching. Afterward, extensive applications performed on public benchmarks for visual retrieval show that using DCT intrinsic orientation achieves performance on a par with SIFT, but with only 60% of its features; replacing the SIFT description with DIFT reduces dimensions from 128 to 32 and improves precision. Image reconstruction resulting from DIFT is presented to show another of its advantages over SIFT.",
author = "Yunhe Wang and Miaojing Shi and Shan You and Chao Xu",
year = "2016",
month = sep,
doi = "10.1109/TIP.2016.2590323",
language = "English",
volume = "25",
pages = "4406--4420",
journal = "IEEE TRANSACTIONS ON IMAGE PROCESSING",
issn = "1057-7149",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "9",

}

RIS (suitable for import to EndNote) Download

TY - JOUR

T1 - DCT-Inspired Feature Transform for Image Retrieval and Reconstruction

AU - Wang, Yunhe

AU - Shi, Miaojing

AU - You, Shan

AU - Xu, Chao

PY - 2016/9

Y1 - 2016/9

N2 - Scale invariant feature transform (SIFT) is effective for representing images in computer vision tasks, as one of the most resistant feature descriptions to common image deformations. However, two issues should be addressed: first, feature description based on gradient accumulation is not compact and contains redundancies; second, multiple orientations are often extracted from one local region and therefore produce multiple descriptions, which is not good for memory efficiency. To resolve these two issues, this paper introduces a novel method to determine the dominant orientation for multiple-orientation cases, named discrete cosine transform (DCT) intrinsic orientation, and a new DCT inspired feature transform (DIFT). In each local region, it first computes a unique DCT intrinsic orientation via DCT matrix and rotates the region accordingly, and then describes the rotated region with partial DCT matrix coefficients to produce an optimized low-dimensional descriptor. We test the accuracy and robustness of DIFT on real image matching. Afterward, extensive applications performed on public benchmarks for visual retrieval show that using DCT intrinsic orientation achieves performance on a par with SIFT, but with only 60% of its features; replacing the SIFT description with DIFT reduces dimensions from 128 to 32 and improves precision. Image reconstruction resulting from DIFT is presented to show another of its advantages over SIFT.

AB - Scale invariant feature transform (SIFT) is effective for representing images in computer vision tasks, as one of the most resistant feature descriptions to common image deformations. However, two issues should be addressed: first, feature description based on gradient accumulation is not compact and contains redundancies; second, multiple orientations are often extracted from one local region and therefore produce multiple descriptions, which is not good for memory efficiency. To resolve these two issues, this paper introduces a novel method to determine the dominant orientation for multiple-orientation cases, named discrete cosine transform (DCT) intrinsic orientation, and a new DCT inspired feature transform (DIFT). In each local region, it first computes a unique DCT intrinsic orientation via DCT matrix and rotates the region accordingly, and then describes the rotated region with partial DCT matrix coefficients to produce an optimized low-dimensional descriptor. We test the accuracy and robustness of DIFT on real image matching. Afterward, extensive applications performed on public benchmarks for visual retrieval show that using DCT intrinsic orientation achieves performance on a par with SIFT, but with only 60% of its features; replacing the SIFT description with DIFT reduces dimensions from 128 to 32 and improves precision. Image reconstruction resulting from DIFT is presented to show another of its advantages over SIFT.

U2 - 10.1109/TIP.2016.2590323

DO - 10.1109/TIP.2016.2590323

M3 - Article

VL - 25

SP - 4406

EP - 4420

JO - IEEE TRANSACTIONS ON IMAGE PROCESSING

JF - IEEE TRANSACTIONS ON IMAGE PROCESSING

SN - 1057-7149

IS - 9

ER -

View graph of relations

© 2020 King's College London | Strand | London WC2R 2LS | England | United Kingdom | Tel +44 (0)20 7836 5454