Interwoven texture-based description of interest points in images

作者：

Highlights：

• This study introduces a formal framework comprised of three principles for construction of feature descriptors in images.

• We propose a novel interwoven texture-based feature descriptor, abbreviated as InterTex, for feature point description in images.

• We validate extensively the performance of InterTex and compare it with the state-of-the-art on multiple datasets and pipelines to show its superiority and robustness under various geometric changes, distortions and noise.

• The proposed InterTex is one of the fastest feature descriptors that can successfully deal with noisy, distorted and texture-limited images.

摘要

•This study introduces a formal framework comprised of three principles for construction of feature descriptors in images.•We propose a novel interwoven texture-based feature descriptor, abbreviated as InterTex, for feature point description in images.•We validate extensively the performance of InterTex and compare it with the state-of-the-art on multiple datasets and pipelines to show its superiority and robustness under various geometric changes, distortions and noise.•The proposed InterTex is one of the fastest feature descriptors that can successfully deal with noisy, distorted and texture-limited images.

论文关键词：Interest point,Descriptive signature,Interwoven texture,Locality,Globality,Robustness

论文评审过程：Received 30 June 2020, Revised 19 December 2020, Accepted 7 January 2021, Available online 16 January 2021, Version of Record 21 January 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.107821