I'm gonna write down the memo of study "Multiple View Geometry in computer vision".
Day 1
Having read the foreword of this book, this sentence was so interesting.
"No designers of systems using cameras hooked to computers that will be built in the foreseeable future can ignore this work"
The contents of the introduction had a summary of this book from 1-1 to 1-10. When I finish reading this, I'll read and memo about this contents.
There are some important keywords that we should learn in each chapters.
Chapter 2. - basic concepts of 2D geometry (recovery of affine properties, metric properties from a perspective img)
Chapter 3. - basic concepts of 3D geometry (plane at infinity, absolute conic)
Chapter 4. - Estimation of geometry from image measurement (minimizing cost function with algebric or geometric measures, robust estimation)
Chapter 5. - How the result of estimation of algorithms may be evaluated
If you ever have any chance(opportunity) to study the geometry of computer vision, I guess you've ever heard that there is some terminology like "Homogeneous coordinates".
Q. "Why do we have to know it? and Why do we use it?
Why do we have to know these concepts?
A. There are a lot of profits for using it for representing non-linear mapping (such as perspective projection, It can be represented by simple matrics multiplications) and point at infinity. point at infinity is the cross point among parallel lines. I'll cover it later.
