Estimating the perceived quality of printed patterns is a complex task as quality is subjective. A study was conducted to evaluate how accurately a machine learning method can predict human judgment about printed pattern quality.

The project was executed in two phases: a subjective test to evaluate the printed pattern quality and development of the machine learning classifier-based automated objective model. In the subjective experiment, human observers ranked overall visual quality. Object quality was compared based on a normalized scoring scale. There was a high correlation between subjective evaluation ratings of objects with similar defects. Observers found the contrast of the outer edge of the printed pattern to be the best distinguishing feature for determining the quality of object.

In the second phase, the contrast of the outer print pattern was extracted by flat-fielding, cropping, segmentation, unwrapping and an affine transformation. Standard deviation and root mean square (RMS) metrics of the processed outer ring were selected as feature vectors to a Support Vector Machine classifier, which was then run with optimized parameters. The final objective model had an accuracy of 83%. The RMS metric was found to be more effective for object quality identification than the standard deviation. There was no appreciable difference in using RGB data of the pattern as a whole versus using red, green and blue separately in terms of classification accuracy.

Although contrast of the printed patterns was found to be an important feature, other features may improve the prediction accuracy of the model. In addition, advanced deep learning techniques and larger subjective datasets may improve the accuracy of the current objective model.

Library of Congress Subject Headings

Imaging systems--Image quality; Machine learning; Three-dimensional printing--Quality control

Publication Date


Document Type


Student Type


Degree Name

Imaging Science (MS)

Department, Program, or Center

Chester F. Carlson Center for Imaging Science (COS)


Jeff B. Pelz

Advisor/Committee Member

Susan Farnand

Advisor/Committee Member

Gabriel Diaz


Physical copy available from RIT's Wallace Library at TK8315 .B37 2017


RIT – Main Campus

Plan Codes