What is the common practice for preprocessing color images with ZCA (Zero Component Analysis):
1) reshape and image and channels into a single feature vector and ZCA the total space?
or
2) apply individual ZCA projections to each channel?

Did anybody experiment with both and is willing to share experiences?

asked Jun 01 '14 at 11:26

eder's gravatar image

eder
2162511

Be the first one to answer this question!
toggle preview

powered by OSQA

User submitted content is under Creative Commons: Attribution - Share Alike; Other things copyright (C) 2010, MetaOptimize LLC.