Understanding and Improving Kernel Local Descriptors

2018/11/27 by Arun Mukundan, Giorgos Tolias, Mukundan, Arun +8 · 1 citation
Computer Science · Engineering · Mathematics · #Advanced Image and Video Retrieval Techniques #Advanced Neural Network Applications #Artificial intelligence #Cartesian coordinate system #Computer Vision and Pattern Recognition (cs.CV) #Computer science #FOS: Computer and information sciences #Kernel (algebra) #Mathematics #Parametrization (atmospheric modeling) #Pattern recognition (psychology) #Pixel #Robotics and Sensor-Based Localization #Robustness (evolution) #cs.CV

paper · pdf · doi:10.48550/arxiv.1811.11147

published in arXiv (Cornell University) (Cornell University)

arxiv created 2018/11/27 · openalex publication_date 2018/11/27 · arxiv updated 2018/11/28 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/04

Abstract

We propose a multiple-kernel local-patch descriptor based on efficient match kernels from pixel gradients. It combines two parametrizations of gradient position and direction, each parametrization provides robustness to a different type of patch mis-registration: polar parametrization for noise in the patch dominant orientation detection, Cartesian for imprecise location of the feature point. Combined with whitening of the descriptor space, that is learned with or without supervision, the performance is significantly improved. We analyze the effect of the whitening on patch similarity and demonstrate its semantic meaning. Our unsupervised variant is the best performing descriptor constructed without the need of labeled data. Despite the simplicity of the proposed descriptor, it competes well with deep learning approaches on a number of different tasks.

Citations

• Understanding Neural Networks Through Deep Visualization
• Aggregating Deep Convolutional Features for Image Retrieval
• A well-conditioned estimator for large-dimensional covariance matrices
• Distinctive Image Features from Scale-Invariant Keypoints
• Multiresolution gray-scale and rotation invariant texture classification with local binary patterns
• Working hard to know your neighbor's margins: Local descriptor learning loss
• Learning to Compare Image Patches via Convolutional Neural Networks
• LIFT: Learned Invariant Feature Transform

Cited by

• Understanding and Optimizing Attention-Based Sparse Matching for Diverse Local Features

Related

• Multiple-Kernel Local-Patch Descriptor
• Beyond Cartesian Representations for Local Descriptors
• Explicit Spatial Encoding for Deep Local Descriptors
• A Large Dataset for Improving Patch Matching
• Feature Fusion for Robust Patch Matching With Compact Binary Descriptors