The Unification of Block Truncation and Transform Coding

Event Date: December 13, 2018
Speaker: Gregory W. Cook
Speaker Affiliation: Samsung Display America Laboratory
Time: 10:00 am
Location: WANG 1004
Priority: No
School or Program: Electrical and Computer Engineering
College Calendar: Show

Gregory W. Cook
Samsung Display America Laboratory

Abstract
Since block truncation coding (BTC) was invented by Delp and Mitchell in 1979 it has been an open problem to create an efficient hybrid of BTC and transform coding.   As observed by Delp and Mitchell in their seminal paper, "...BTC does produce sharp edges; however, these edges do have a tendency to be ragged.  Transform coding usually produces edges that are blurred and smooth."   Consequently, the best hybrid coding algorithm would utilize the best features of both algorithms.

At the heart of the issue is the fact that BTC produces an arbitrary mask, and any transform which *follows* BTC must operate efficiently on this arbitrary mask.  The Sparse Transform (ST) exactly fills this requirement. Initially developed for the discrete cosine transform (DCT), the ST has been extended to use a set of orthogonal transforms which include the DCT, Walsh-Hadamard, Slant, High-correlation Transform (HCT, used in AVC/H.264), and the Harr Wavelet transform. The ST has the properties of (a) operating on an arbitrary pixel mask (b) perfectly invertible  (c) transform coefficients in the vector space of the largest transform and (d) extendible to any size transform with length equal to a power of two.

This talk will develop the basics of BTC, the basics of the ST, and show the unified coding algorithm as a cascade of the two methods. Preliminary results as applied to arbitrary shaped blocks for object coding will be shown to demonstrate the efficacy of the method. 

Bio
Gregory W. Cook received the B.E.E. (with highest honors) and the M.S.E.E. degrees from the Georgia Institute of Technology, Atlanta, in 1984 and 1985, respectively, and the Ph.D. degree from Purdue
University, West Lafayette, Indiana, in 2002.  From 1985 to 1991, he was a commissioned officer in the United States Air Force, first working at the Avionics Laboratory and then as an Instructor of Electrical Engineering at the United States Air Force Academy. In 2003, he was a Postdoctoral Research Associate at the Indiana Center for Biological Microscopy.  Starting in 2004,  Dr. Cook worked as a
Senior Staff Research Engineer at Technicolor, Futurewei, HP Labs, and is currently at the Samsung Display America Lab in San Jose, California.  His research interests include image and video
compression, wireless transmission, three dimensional data display, and objective image quality metrics.  He holds nine US patents and has over 25 journal and conference publications.  Dr. Cook is a 1993-1994 recipient of an Intel Foundation Graduate Fellowship, and is a member of the Institute for Electrical  and Electronics Engineers and the Society for Information Display. 

Host
Edward J. Delp

2018-12-13 10:00:00 2018-12-13 11:00:00 America/Indiana/Indianapolis The Unification of Block Truncation and Transform Coding WANG 1004