Abstract:Local feature extraction plays a crucial role in numerous critical visual tasks. However, there remains room for improvement in both descriptors and keypoints, particularly regarding the discriminative power of descriptors and the localization precision of keypoints. To address these challenges, this study introduces a novel local feature extraction pipeline named OSDFeat (Object and Spatial Discrimination Feature). OSDFeat employs a decoupling strategy, training descriptor and detection networks independently. Inspired by semantic correspondence, we propose an Object and Spatial Discrimination ResUNet (OSD-ResUNet). OSD-ResUNet captures features from the feature map that differentiate object appearance and spatial context, thus enhancing descriptor performance. To further improve the discriminative capability of descriptors, we propose a Discrimination Information Retained Normalization module (DIRN). DIRN complementarily integrates spatial-wise normalization and channel-wise normalization, yielding descriptors that are more distinguishable and informative. In the detection network, we propose a Cross Saliency Pooling module (CSP). CSP employs a cross-shaped kernel to aggregate long-range context in both vertical and horizontal dimensions. By enhancing the saliency of keypoints, CSP enables the detection network to effectively utilize descriptor information and achieve more precise localization of keypoints. Compared to the previous best local feature extraction methods, OSDFeat achieves Mean Matching Accuracy of 79.4% in local feature matching task, improving by 1.9% and achieving state-of-the-art results. Additionally, OSDFeat achieves competitive results in Visual Localization and 3D Reconstruction. The results of this study indicate that object and spatial discrimination can improve the accuracy and robustness of local feature, even in challenging environments. The code is available at https://github.com/pandaandyy/OSDFeat.

Localizing Artistic Objects in Frescos Through Contour Based Features

Style strokes extraction based on color and shape information

Real-time browsing of high-precision frescos

A computer-aided style stroke extraction system for "Dun Huang" frescoes

FSIS: Fast and Seamless Image Stitching of High-Resolution Dunhuang Murals

Aesthetic Visual Style Assessment on Dunhuang Murals

Similarity metrics between mural images with constraints of the overall structure of contours

Content And Semantic-Based Image Retrieval For The Preservation, Restoration And Pattern Creation Of Dunhuang Fresco

Knowledge Based Lacunas Detection And Segmentation For Ancient Paintings

A Method for Accurate Localization of Facial Features

FreSCo: Frequency-Domain Scan Context for LiDAR-based Place Recognition with Translation and Rotation Invariance

Dunhuang Mural Line Drawing Based on Multi-scale Feature Fusion and Sharp Edge Learning

Color restoration and image retrieval for Dunhuang fresco preservation

Dunhuang murals contour generation network based on convolution and self-attention fusion

Constructing a mobile visual search framework for Dunhuang murals based on fine-tuned CNN and ontology semantic distance

CF-lines: a Fusing Contour Features Optimization Method for Line Segment Detector

A Local Shape Descriptor for Mobile Linedrawing Retrieval

Object and spatial discrimination makes weakly supervised local feature better

Exploring the topic evolution of Dunhuang murals through image classification

Object Retrieval and Localization in Large Art Collections using Deep Multi-Style Feature Fusion and Iterative Voting

Localization, balance and affinity: a stronger multifaceted collaborative salient object detector in remote sensing images