Table of contents
2.2.3 Morphological operations
2.2.4 Spatial domain processing and transformation
2.2.5 Spatial domain analysis and transformation
(4) Gradient Prewitt filtering
(6) Gradient Laplacian filtering
2.2.6 Frequency domain analysis and transformation
2. Image preprocessing
See courseware for details:
2.1 Introduction
Image preprocessing is actually an image enhancement process:
Spatial domain:
Point operation: It is to adjust the overall color difference of the image based on the histogram, and adjust the color of a point, which is more or less related to the surroundings
Morphological operations: corrosion, expansion
Local operation: Each point is compared with its surrounding points or calculated together
Frequency domain:
Mapping the spatial domain to the frequency domain, for us here, the meaning isFast calculation of convolution
Fuli leaf change, small wave calculation
2.2 Feature extraction method
2.2.1 Histogram
2.2.2 CLASH
2.2.3 Morphological operations
2.2.4 Spatial domain processing and transformation
2.2.5 Spatial domain analysis and transformation
1. Convolution
The most commonly used method is zero padding, but it still has a great impact on the calculation results. Therefore, it is necessary to avoid using large convolution kernels to convolve relatively small images. Please refer to the courseware for specific strategies.
2. Filtering
(1) Mean filtering
(2) Median filtering
(3) Gaussian filter
It can be used in Gaussian pyramid. When the picture is far away from you, you can keep those pixels to make this picture closest to the original picture in a certain sense.
(4) Gradient Prewitt filtering
(5) Gradient Sobel filtering
(6) Gradient Laplacian filtering
Laplacian, sum = 0, avoid selecting smooth areas
Original image minus Laplacian filter
(7) Other filtering
2.2.6 Frequency domain analysis and transformation
(1) Gaussian Pyramid
(2) Laplace’s Pyramid
(3) Fourier transform
(4) Wavelet transform