Computer vision tasks {computer vision tasks} include content-based image retrieval, depth perception, egomotion, event detection, identification, image restoration, indexing, learning, object recognition, pose estimation, scene reconstruction, and tracking.
art gallery problem
Find minimum-length path through art gallery from which one can see all pictures. Watchman-route problem, safari problem, zookeeper problem, touring-polygons problem, and parts-cutting problem have same type. Rubberband algorithms solve them.
barcode reading
Decode 1D and 2D codes.
character recognition
Recognize serial numbers, words, and phrases.
content-based image retrieval
In image sets, find images with content, such as text, number, object, or image.
dense stereo vision
Two cameras, with known or unknown separation and angle, can find scene-point depths.
depth perception
One eye can use linear perspective, motion parallax, interposition, shading, relative size, relative height, aerial perspective, texture, and three-dimensional-structure motion. Two eyes can use convergence and binocular disparity.
egomotion
Calculate camera three-dimensional motion.
event detection
Find abnormal or special feature or property in images.
gauging
Measure object dimensions.
human-machine interface
Algorithms allow human and robot interaction.
identification
Match individual image to stored image.
image restoration
Remove noise using low-pass filters, median filters, or image models.
kinematic chain
Rigid bars connect by sliding or rotating joints.
object recognition
Recognize learned object type at image location.
optical flow
When camera or person moves, scene flows past. Lucas-Kanade method, Horn-Schunck method, Nagel-Enkelmann method, correlation method, and block-matching method use variational methods to find optical flow.
pose estimation
Find object position or orientation.
scene reconstruction
Using several scene images, calculate three-dimensional model.
structure from motion
Motions cause disparities and disparity rates, which can reveal structure. Bundle-adjustment algorithms can find three-dimensional scene structure and camera trajectories.
First, projective reconstruction can construct projected structure, then Euclidean upgrading can find actual shape. Affine reconstruction uses Tomasi-Kanade factorization.
template matching
Find, match, and/or count patterns.
tracking
Follow object velocities and directions.
Mathematical Sciences>Computer Science>Systems>Computer Vision>Tasks
3-Computer Science-Systems-Computer Vision-Tasks
Outline of Knowledge Database Home Page
Description of Outline of Knowledge Database
Date Modified: 2022.0224