Abstract
Background
Image descriptors are an ubiquitous tool in computer vision. By representing the content of an image or an image region in a compact and robust way, they make matching problems more efficient, as shown in Figs. 1 and 2. Typically, a descriptor is used to query a set of descriptors, looking for the most similar descriptor in the set. Efficient algorithms, such as hashing, can be used to make this search extremely fast, even for large databases, when the Euclidean distance can be used as similarity measure. Applications range from simultaneous localization and mapping (SLAM) and Structure from Motion (SfM) to image retrieval and object recognition.
Many different approaches were proposed over the years, and as almost any computer vision topic, deep learning has also changed the way descriptors could be computed. We describe below the most representative approaches to computing descriptors, as …
Image descriptors are an ubiquitous tool in computer vision. By representing the content of an image or an image region in a compact and robust way, they make matching problems more efficient, as shown in Figs. 1 and 2. Typically, a descriptor is used to query a set of descriptors, looking for the most similar descriptor in the set. Efficient algorithms, such as hashing, can be used to make this search extremely fast, even for large databases, when the Euclidean distance can be used as similarity measure. Applications range from simultaneous localization and mapping (SLAM) and Structure from Motion (SfM) to image retrieval and object recognition.
Many different approaches were proposed over the years, and as almost any computer vision topic, deep learning has also changed the way descriptors could be computed. We describe below the most representative approaches to computing descriptors, as …
Original language | English |
---|---|
Title of host publication | Computer Vision: A Reference Guide |
Publisher | Springer International |
Pages | 1-8 |
Publication status | Published - 2020 |