A computationally efficient method to compare the shape of planar Gaussian mixtures using their underlying distribution of distances and web-based diagnosis tools for customers to self-solve printing issues with electrophotographic printers.

摘要

This dissertation proposes a novel shape matching methodology for objects represented by a planar Gaussian mixture, and describes the design of two web-based troubleshooting tools for printing issues with electrophotographic printers. The motivation of the shape matching methodology is the problem of recognizing planar objects consisting of "blobs" that can be modeled as weighted Gaussian densities (e.g., the halftone patterns in a print). We first describe an empirical comparison method assuming a large number of independent samples are given for each distribution. This recognition method is extended to the case where one Gaussian Mixture is a known template and the other Gaussian mixture consists of an observed sparse set of points (e.g., the minutiae of a fingerprint). Instead of comparing the Gaussian mixtures directly, we compare the underlying distribution of distances of each mixture. Since distances are invariant under rotations and translations, this provides a workaround to the problem of aligning the objects before comparing them---thus speeding the comparison process. We prove that the distribution of distances is a lossless representation of the shape of generic Gaussian mixtures, and show that the proposed method is no less accurate than methods which compare the planar mixtures directly. The remaining discussion focuses on the design of two web-based troubleshooting tools for print quality and printing color issues. Both issues poses special challenges for a manufacturer's support organization, and a quick resolution is an important factor for customer satisfaction. We review the process for developing the websites, and the organization of their content.