Hi,
From the question I understand that you need to find closest image for each image in ‘folder1’ in ‘folder2’. As mentioned in a question answered here the similarity can be established between two images structurally. So, for each pair of images you may run ssim and get the similarity index stored in a matrix where rows represent images in ‘folder1’ and column represent images in ‘folder2’. The maximum value in a row corresponds to most similar image based on the column number. Since you have mentioned that image is (64x64) the ssim will work as it support grayscale images as of now.
