Zhentao-Liu / KLT-JND

This is a brief introduction of our proposed JND profile with totally new design philosophy based on KLT.
27 stars 1 forks source link

KLT Decomposition question #2

Closed hbrachemi closed 1 year ago

hbrachemi commented 1 year ago

Hey, I might have few questions regarding the code; When computing the KLT coefficients:

KLT_kernel = pca(text_matrix'); klt_coeff = KLT_kernel'*text_matrix;

the pca function normally returns the klt coefficients, I don't seem to understand why you used those coefficients as a kernel and had to multiply them later by the text matrix; Thanks in advance

hbrachemi commented 1 year ago

Figured out (pca(a',Centered=false)'*a)' in matlab is equivalent to pca.fit_transform in python thus the pca function only returns the kernel and not the projection into the new space