[][src]Crate cv


Batteries-included pure-Rust computer vision crate

All of the basic computer vision types are included in the root of the crate. Modules are created to store algorithms and data structures which may or may not be used. Almost all of the things in these modules come from optional libraries. These modules comprise the core functionality required to perform computer vision tasks.

Some crates are re-exported to ensure that you can use the same version of the crate that cv is using.


  • camera - camera models to convert image coordinates into bearings (and back)
  • consensus - finding the best estimated model from noisy data
  • geom - computational geometry algorithms used in computer vision
  • estimate - estimation of models from data
  • feature - feature extraction and description
  • knn - searching for nearest neighbors in small or large datasets
  • optimize - optimizing models to best fit the data



Camera models


Consensus algorithms


Estimation algorithms


Feature detection and description algorithms


Computational geometry algorithms


Algorithms for performing k-NN searches




Optimization algorithms




A constant sized array of bits. B defines the number of bytes. This has an alignment of 64 to maximize the efficiency of SIMD operations. It will automatically utilize SIMD at runtime where possible.


A 3d point which is relative to the camera's optical center and orientation where the positive X axis is right, positive Y axis is down, and positive Z axis is forwards from the optical center of the camera. The unit of distance of a CameraPoint is unspecified and relative to the current reconstruction.


This contains a camera pose, which is a pose of the camera relative to the world. This transforms camera points (with depth as z) into world coordinates. This also tells you where the camera is located and oriented in the world.


This stores an essential matrix, which is satisfied by the following constraint:


Normalized keypoint match


Normalized keypoint to world point match


A point on an image frame. This type should only be used when the point location is on the image frame in pixel coordinates. This means the keypoint is neither undistorted nor normalized.


Allows pose solving to be parameterized if defaults are not sufficient.


This contains a relative pose, which is a pose that transforms the CameraPoint of one image into the corresponding CameraPoint of another image. This transforms the point from the camera space of camera A to camera B.


Contains a member of the lie algebra so(3), a representation of the tangent space of 3d rotation. This is also known as the lie algebra of the 3d rotation group SO(3).


This stores a RelativeCameraPose that has not had its translation scaled.


A point in "world" coordinates. This means that the real-world units of the pose are unknown, but the unit of distance and orientation are the same as the current reconstruction.


This contains a world pose, which is a pose of the world relative to the camera. This maps WorldPoint into CameraPoint, changing an absolute position into a vector relative to the camera.



Describes the direction that the projection onto the camera's optical center came from. It is implemented on projection items from different camera models. It is also implemented for Unit<Vector3<f64>> if you want to pre-compute the normalized bearings for efficiency or to turn all camera models into a unified type.


Allows conversion between the point on an image and the internal projection which can describe the bearing of the projection out of the camera.


A consensus algorithm extracts a consensus from an underlying model of data. This consensus includes a model of the data and which datapoints fit the model.


An Estimator is able to create a model that best fits a set of data. It is also able to determine the residual error each data point contributes in relation to the model.


Allows the retrieval of the point on the image the feature came from.


This trait is implemented by points inside of a metric space.


A model is a best-fit of at least some of the underlying data. You can compute residuals in respect to the model.


See Consensus. A multi-consensus can handle situations where different subsets of the data are consistent with different models. This kind of consensus also considers whether a point is part of another orthoganal model that is known before assuming it is a true outlier. In this situation there are inliers of different models and then true outliers that are actual erroneous data that should be filtered out.


This trait is for algorithms which allow you to triangulate a point from two or more observances. Each observance is a CameraPose and a Bearing.


This trait allows you to project the point a onto the bearing b by only scaling a translation vector t. The returned value is the amout to scale t by to achieve the triangulation. All inputs share the same origin (optical center). Below is a visualization of the problem.


This trait allows you to take one relative pose from camera A to camera B and two bearings a and b from their respective cameras to triangulate a point from the perspective of camera A.