event

Center for Signal and Information Processing (CSIP) Seminar

Primary tabs

Speaker: Thrasyvoulos (Thrasos) N. Pappas, Professor, Electrical Engineering and Computer Science Department, Northwestern University, Evanston, Illinois

Title: Visual Texture Analysis: From Similarity to Material Properties

Abstract:
Texture is an important visual attribute for both human perception and image analysis. It provides important clues for object shape and boundary detection, as well as material identification. Our research has initially focused on texture similarity, which is important for a variety of applications, including image compression and content-based retrieval. We have proposed a new class of structural texture similarity metrics (STSIMs) that account for human visual perception and the stochastic nature of textures. They rely entirely on local image statistics and allow substantial point-by-point deviations between textures that according to human judgment are similar or essentially identical. We have also developed new testing procedures for objective texture similarity metrics. We have identified three operating domains for evaluating the performance of such metrics, each with different performance goals and testing procedures. We have also proposed ViSiProG (Visual Similarity by Progressive Grouping), a new procedure for collecting subjective similarity data.

Our current goal is the extraction of material properties.  We propose techniques that account for dramatic changes in texture appearance due to variations in illumination and viewing conditions.  The key elements the proposed approach are (1) the use of ViSiProG for identifying clusters of visually similar textures; (2) the characterization of each material with a small set of exemplars; (3) the use of machine learning techniques for training STSIMs to separate the textures according to the material they correspond to.  We demonstrate the effectiveness of the proposed techniques using "CUReT," a fully labeled database of real-world surfaces, viewed under different illuminations and viewing angles at a fixed distance.  However, the proposed approaches can be applied to domains where semantic labeling is not available, for example, a database of satellite images, and a database of building fronts obtained from "Google Earth Street View.”

Speakers Bio:
Thrasos Pappas received the Ph.D. degree in electrical engineering and computer science from MIT in 1987. From 1987 until 1999, he was a Member of the Technical Staff at Bell Laboratories, Murray Hill, NJ. He joined the EECS Department at Northwestern in 1999.  His research interests are in human perception and electronic media, and in particular, image and video quality and compression, image and video analysis, content-based retrieval, model-based halftoning, and tactile and multimodal interfaces. Prof. Pappas is a Fellow of the IEEE, SPIE and IS&T.  He has served as editor-in-chief of the IEEE Transactions on Image Processing (2010-12), elected member of the Board of Governors of the Signal Processing Society of IEEE (2004-07), chair of the IEEE Image and Multidimensional Signal Processing Technical Committee (2002-03), and technical program co-chair of ICIP-01 and ICIP-09.  He has also served as VP-Publications for the Signal Processing Society of IEEE. From 1997 to 2017, he has served as co-chair of the SPIE/IS&T Conference on Human Vision and Electronic Imaging. He is currently a founding Editor-in-Chief of the IS&T Journal of Perceptual Imaging.

Status

  • Workflow Status:Published
  • Created By:Ashlee Gardner
  • Created:04/11/2018
  • Modified By:Ashlee Gardner
  • Modified:04/11/2018

Keywords

  • No keywords were submitted.