RTV in Python

Relative Total Variation(a method for structure extraction from texture)

Related paper can be found in http://www.cse.cuhk.edu.hk/leojia/projects/texturesep/texturesep12.pdf

it is hard to say,but i think this method is not very impressive and efficient.

this is how RTV works in images with textures.

the corresponding MatLab codes and demo images can be found in folder "matlab code"

This is the Python Version of RTV and there may be some small problems such as the memory related error. Code line 232, A = A.A ,when i try to convert scipy.sparse.csr.csr_matrix to numpy.ndarray ,if the image is large( specifically, i think over 80*80 is large enough) ,a MemoryError is raised.

i test the method on lena.

this is the result of max_iter=4 and img-size 64*64:

this is the result of max_iter=2 and img-size 64*64:

you can test MatLab codes on https://octave-online.net/ ,it is a convenient matlab online tool. but the large-image-problem still occurs in another way: out of time. :)