Nowadays in the field of computer vision, convolution is probably the most frequently appearing term. This is because convolutional neural networks are widely used to solve various computer vision tasks and have become a default tool in the field of computer vision. However, the term convolution does not come from the field of computer science, but from the field of signal processing. Convolution is an important operation in signal processing, used to calculate the output response of a linear system to an input signal. What is convolution? As the name suggests, it is to “flip” first, then “accumulate”. In professional terms, the input of convolution is two functions, one is the input signal function, the other is the system function; the convolution operation first flips one function, then slides it over the other function for accumulation, and finally gets the output result, as shown in Fig. 2.1. In the case of continuous functions, superposition refers to the integration of the product of two functions, while in the discrete case it is the weighted sum of functions, for simplicity, it is collectively referred to as superposition.