hacarus / spm-image

Sparse modeling and Compressive sensing in Python
105 stars 29 forks source link

improve ksvd efficiency #103

Open sutkss opened 3 years ago

sutkss commented 3 years ago

Issue Description

The result of KSVD algorithm may contain atoms which has 0 coefficient for all samples. The cause is due to: https://github.com/hacarus/spm-image/blob/development/spmimage/decomposition/ksvd.py#L88-L89

Replace such all 0 coef atoms with right-singular vectors of SVD for error Y - WH.