SIFT (1)

There are many SIFT blogs either on cnnblog or csdn. I reviewed many. However, in my view, there are not in details. Mainly that they 

do not offer a good feeling and real touching to the fundamentals underlying the math.

So I try to present the direct feeling and real touches.

 

2D feature extracted, satisfy:

1. scale invarient;

2. rotation invarient.

 

Naive Thinking: Minimal point or Maximum Point, Local;

Now you are given a grey img. Just a Matrix with element value range from 0,1,.., to255.

It is a finite discrete set.

 

The details looks like this: