Correlation or Convolution during Local Frequency Analysis

[DeckerDai]

Hi Cattle, thanks for your wonderful code!

I notice that you use imfilter in gabor_gradient_field.m. The default operation adopted by imfilter is correlation, instead of convolution. However, the operation in the Paper is said to be convolution.

Theoretically, I believe correlation is more suitable here, since we want to identify which filter (corresponds to a frequency) provides a higher response. What do you think?

Thank you and look forward to hearing from you.

Best regards, Qi