cross-correlation和convolution

.

python裏，互相關的函數。scipy.signal.correlate(in1, in2, mode='full')

correlate(in1, in2, mode='full')
    Cross-correlate two N-dimensional arrays.
    
    Cross-correlate `in1` and `in2`, with the output size determined by the
    `mode` argument.
    
    Parameters
    ----------
    in1 : array_like
        First input.
    in2 : array_like
        Second input. Should have the same number of dimensions as `in1`;
        if sizes of `in1` and `in2` are not equal then `in1` has to be the
        larger array.
    mode : str {'full', 'valid', 'same'}, optional
        A string indicating the size of the output:
    
        ``full``
           The output is the full discrete linear cross-correlation
           of the inputs. (Default)
        ``valid``
           The output consists only of those elements that do not
           rely on the zero-padding.
        ``same``
           The output is the same size as `in1`, centered
           with respect to the 'full' output.
    
    Returns
    -------
    correlate : array
        An N-dimensional array containing a subset of the discrete linear
        cross-correlation of `in1` with `in2`.
    
    Notes
    -----
    The correlation z of two arrays x and y of rank d is defined as:
    
      z[...,k,...] = sum[..., i_l, ...]
                         x[..., i_l,...] * conj(y[..., i_l + k,...])

 

參考 https://www.zhihu.com/question/29121110