Simply, correlation is a measure of similarity between two signals, and convolution is a measure of the effect of one signal on the other.
在图像或者信号处理领域,简单说,correlation 用来评估两个信号的相似度,而 convolution 是用来评估一个信号对另一信号影响的度量。
在数学上,可以说 convolution = folding + correlation,即 correlation 的 kernel 逆时针翻转 180 度,再和输入进行 correlation 操作就是 convolution。具体可以看下面代码(没有进行 padding 操作):