Intelligent Computer Vision 1(개요)
목차
수업교재
Class Materials
- Multiple view geometry in computer vision (2nd)
- Computer Vision: Algorithms and Applications (2010)
Computer Vision
Szeliski p.1~28
A theme in the development of this field has been to duplicate the abilites of human vision by electronically perceiving and understanding an image.
Goal of computer vision
=> Write computer programs that can interpret images.
Can computers match human vision
=> Mostly no
- humans are much better at "hard" things
- computers can be better at "easy" things
Current state of the art (Application Area)
Univ. Washington cse455 2012
- Earth Viewers (3D modeling)
- Motion Capture
- Optical Character Recognition (OCR)
- Face detection
- Smile detection
- Object recognition
- Face recognition
- Motion capture
- Sports
- Smart cars
- Self-driving cars
- Robotics
- Vision in space
- Medical imaging
Image Formation
E(x, y, z, lambda) ※ 좌표 + 빛의 파장
Camera(x, y, z, lambda) = E * R
E: Incident light R: Reflectivity function
Projection
Camera_projection(x', y', lambda) = Projection(Camera)
Perspective: Pinhole
=> Pinhole effect
big: blurring
small: diffractionOrthogonal
Sensitivity
V_2(lambda) - lambda
Summary
The Image function f_c(x', y') (C = R, G, B) is formed as:
f_c(x', y') = Int{ Camera_projection * V_c(lambda) } delta lambda
Degital Image Formation
We have to discretize:
- x', y' => x'_i, y'_j => Sampling
- f_c(x'_i, y'_j) ∈ Range => Quantization
Quantization to P levels
Typically P = 2^8 = 256 (8bit quantization)
Color Depth
- Monochrome (1bit)
- Gray-scale (8bits)
- Color (8 or 16 bits)
- Ture color (24 or 32 bits)
RGB (24 bits)
RGB + Alpha (32bits)
Color Space
- Humans (Hue Saturation Lightness/Brightness, HSL or HSB)
- CRT monitors (RGB) *additive
- Printers (Cyan Magenta Yellow Black, CMYK) *subtractive
- Compression (Luminance and Chrominance, YIQ YUV YChCr)
Y = 0.3R + 0.59G + 0.11B
U = (B-Y) 0.493
V = (R-Y) 0.877
Visible Spectrum
UV(400nm) ~ IR (700nm)
Tips
== Classical Vision ==
- Overview
- Filtering (Edge detection)
- Keypoint detection (Corner detection, SIFT ..)
- Hough transform
== Multiple View Geometry == - 2D geometry
- 3D geometry
- Camera Models and Calibration
== Deep Learnings == - Image Classification (BoW)
- Object detection (VJ)
- Image Segmentation (Graph-cuts)
- Optical flow (Lukas-Kanade)
- Pose estimation (DPM)
- 3D Point cloud
Reference
2021 MIPAL, Gwak Nojun