Motivation for studying Computer Vision
An image captures a 2D projection from the 3D world. The projection
process inherently involves some information loss. Despite this
humans and animals are surprisingy skilled at acting and reasoning
based on visual information.
Computer Vision is the study about how to recover properties of the
3D world from 2D images or video. It is related to computer graphics,
where the (easier) task of rendering 2D images from a 3D model is
dealt with, and image-processing which transforms 2D images into
new (better) 2D images.
On the high level, tasks in Computer Vision are often divided into
object recognition ("what" is in the image) and localization ("where"
something is located). The former is concerned with computing local
properties of a subpart of an image to use as evidence for the presence
of a particular object. The latter involves involves finding 3D geometric
information from the image from outright 3D reconstruction, to using
local measurements for reasoning about alignmnents of controlling the
motions of a robot ("how").
To solve the high level tasks often significant low-level processing is
required for instance feature detection, line finding, segmentation,
computing motion, computing corresponding points.
Computer Vision can be studied for a variety of reasons:
Of scientific interest
- Understand geometry of imaging and camera models.
- Understand how light interacts with surfaces.
- Explore the human vision system and cognition.
Of practical interest
- Industrial inspection.
- Medical imaging: e.g. recover and analyse 3D MRI.
- Movie special effects: e.g. match real world images with graphics.
- Video game: Interact with game using camera.
Capture and insert a 3D model of yourself into multiplayer game.
- Security: Person identification, tracking.
- Mapping: Identify regions in satellite imagery.
- Search: Index and find images in databases based on visual keys.
Of cross-disciplinary interest
- The mathematical fundament of computer vision and graphics is mostly the same.
- In AI researchers make high level inferences from visual data.
- Image processing, Vision and Graphics all share/involve 2D processing.
- In Robotics and Engineering vision is used to control motion and other processes.
- In Neuroscience and Psychology human and animal vision is studied.
- In Design, Art, Architecture digital modeling is increasingy common.
- Last but not least, its a fun and visual ways of learning from fundamental
math and physics to how to write algorithms for 2D and 3D processing.