In order to avoid the text expression is not intuitive enough, the above figure is directly to illustrate the processing method of the spatial domain.
The following describes the methods mentioned above one by one.
Numerical operations
As shown in the following figure, numerical operations can be basically divided into processing of single-channel images and processing of multi-channel images because of the specific operation objects.
Single-channel point operation is to directly process the value of the pixel, such as logarithmic transformation, power-law transformation, piecewise transformation, etc., all apply a mathematical formula to each pixel. The typical and common neighborhood operations are filters, such as mean smoothing, Gaussian smoothing, sobel filter, etc.
set operations
The basic operations of union, intersection and difference between the same image and different images.
For example, image mosaicking is to find the union (overlapping area) of two images and stack them together, and then form a new image. Intuitively, the image becomes larger. As shown below:
logic operation
Logical operations are relatively simple, because they are all logical matrices, and there are only two cases of 0 and 1. It is basically the same as our previous logical operations, except that most of them are in the form of matrices.