Perceptual metrics, visualization tools, and machine-learning-based quality prediction for evaluation of page uniformity

Minh Q Nguyen, Purdue University

Abstract

Uniformity is one of the issues of most critical concern for laser electrophotographic (EP) printers. Typical non-uniformity defects include mottle, grain, pinholes, and finger prints. Among these defects, mottle (low spatial frequency) and grain (high spatial frequency) are the most commonly observed. In order to assess Print Quality (PQ), we propose a novel method that uses a block-based technique to analyze the test page both visually and metrically. With a print-to-scan method, we use a grid of 150 pixels x 150 pixels (1/4 inch x 1/4 inch scanned at 600 dpi) square blocks throughout the scanned page. For each block, we examine two aspects: behavior of its pixels within the block (metric of graininess) and behavior of the blocks within the printed page (metric of uniformity). For an input scanned page, we create eight visual outputs, each displaying a different aspect of uniformity by using various pseudo color codings. We next introduce a set of tools for machine learning in the assessment of printed page uniformity. This work is primarily targeted to the printing industry, specifically the ubiquitous laser, electrophotographic printer. We use selected features that are well-correlated with the rankings of expert observers to develop a novel machine learning framework that allows one to achieve the minimum "false alarm" rate, subject to a chosen "miss" rate. The features are collected from the previous eight models. A "miss" is defined to be a page that is not of acceptable quality to an expert observer that the prediction algorithm declares to be a "pass". Misses are a serious problem, since they represent problems that will not be seen by the systems designers. On the other hand, "false alarms" correspond to pages that an expert observer would declare to be of acceptable quality, but which are flagged by the prediction algorithm as "fails". In a typical printer testing and development scenario, such pages would be examined by an expert, and found to be of acceptable quality after all. "False alarm" pages result in extra pages to be examined by expert observers, which increases labor cost. But "false alarms" are not nearly as catastrophic as "misses", which represent potentially serious problems that are never seen by the systems developers. This scenario motivates us to develop a machine learning framework that will achieve the minimum "false alarm" rate subject to a specified "miss" rate. In order to construct such a set of receiver operating characteristic (ROC) curves, we conduct an exhaustive search over the space of the nonlinear discriminants of the Cost-Sentitive Support Vector Machine (SVM) framework. Our work shows promise for applying a standard framework to obtain a full ROC curve when it comes to tackling other machine learning problems.

Degree

Ph.D.

Advisors

Allebach, Purdue University.

Subject Area

Electrical engineering

COinS