Characterizing Streaks in Printed Images: A Matching Pursuit Method using Wavelet Decomposition

Juan Liu, Wencheng Wu, Bob Price, Eric Hamby, and Raj Minhas
Submission Type: 
Full Paper
phmc_10_078.pdf1.09 MBOctober 12, 2010 - 5:23pm

Printed pages from industrial printers can exhibit a number of defects. One of the common defects and a key driver of maintenance costs is the line streak. This paper describes an efficient streak characterization method for automatically interpreting scanned images using the matching pursuit algorithm. This method progressively finds dominant streaks in signal profiles. It uses wavelet decomposition to speed up the element selection process and reduce computation complexity. Previous approaches require the design engineer to pre-specify the characteristics for each possible streak that could be detected -- an approach which is practically limited to detecting a few streak types in specific locations. The Matching Pursuit algorithm, inc contrast, fully characterizes any and all streaks found on the scanned page permitting a generic analysis of a broad range of defects found in the field.

Publication Control Number: 
Submission Keywords: 
data preprocessing
feature extraction
Automatic diagnostics
Submitted by: 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed