Received: from gauss.eecs.berkeley.edu (gauss.EECS.Berkeley.EDU [128.32.240.14]) by hera.EECS.Berkeley.EDU (8.6.4/8.6.4) with ESMTP id BAA13102; Sat, 5 Feb 1994 01:00:07 -0800 Received: from localhost (shuvra@localhost) by gauss.eecs.berkeley.edu (8.6.4/8.6.4) id BAA16346; Sat, 5 Feb 1994 01:00:05 -0800 Date: Sat, 5 Feb 1994 01:00:05 -0800 From: Shuvra Bhattacharyya Message-Id: <199402050900.BAA16346@gauss.eecs.berkeley.edu> To: 231cory@hera.EECS.Berkeley.EDU, ee298-20@mho.EECS.Berkeley.EDU, seminar@hera.EECS.Berkeley.EDU, ucb-seminars@ucbvax.berkeley.edu Subject: February 10, Special Digital Signal Processing Seminar Status: R Special Digital Signal Processing Seminar 3:00 - 4:00pm Thursday, February 10, 1994 Hughes Room, 400 Cory Hall Stochastic Image Modeling For Image Enhancement Robert L. Stevenson Laboratory for Image and Signal Analysis Department of Electrical Engineering University of Notre Dame Notre Dame, IN 46556 rls@stargate.ece.nd.edu ABSTRACT ________ As a result of their mathematical simplicity, linear operators have long been the primary tool for signal and image processing. With the need for improved performance, increasingly more sophis- ticated algorithms have been devised. Yet despite their many practical applications, there are several important problems which can not be satisfactorily addressed with just linear opera- tors. In situations where the signal of interest has discon- tinuities, a linear operator will tend to smooth such discon- tinuities, leading to lower signal resolution and poorer perfor- mance. Another important situation occurs when the noise en- vironment contains large impulses which corrupt the signal; linear operators are not effective in removing such contamina- tion. This talk will discuss a nonlinear estimation technique for pro- ducing high quality signal estimates. The problem is stated within the framework of maximum a posteriori (MAP) estimation. Appropriate signal and noise models are developed so that stable, high-resolution and robust signal estimates can be computed. Ex- amples will be shown from the area of image processing and novel computational structures for high-speed real-time computation will be discussed. ____________________ Administrative requests regarding this seminar should be sent to dsp-seminar-request@diva.eecs.Berkeley.EDU. Abstracts for the DSP seminar are available by anonymous ftp from ptolemy.eecs.berkeley.edu. They are stored in the directory ~ftp/pub/misc/dsp-seminar-abstracts. From liza@cs.Berkeley.EDU Fri Nov 4 14:54:30 1994 Received: from hofmann.CS.Berkeley.EDU (hofmann.CS.Berkeley.EDU [128.32.34.35]) by hera.EECS.Berkeley.EDU (8.6.9/8.6.9) with ESMTP id OAA15448; Fri, 4 Nov 1994 14:54:26 -0800 Received: from hera.EECS.Berkeley.EDU (hera.EECS.Berkeley.EDU [128.32.240.97]) by hofmann.CS.Berkeley.EDU (8.6.9/8.6.6.Beta11) with ESMTP id OAA11968; Fri, 4 Nov 1994 14:54:08 -0800 Received: (from dbolt@localhost) by hera.EECS.Berkeley.EDU (8.6.9/8.6.9) id OAA15347 for eecsdept; Fri, 4 Nov 1994 14:53:07 -0800 Date: Fri, 4 Nov 1994 14:53:07 -0800 From: Dianna Bolt Message-Id: <199411042253.OAA15347@hera.EECS.Berkeley.EDU> To: eecsdept@hera.EECS.Berkeley.EDU Subject: Strategic Directions Seminars Status: RO What follows is the schedule for the remaining talks to be given in the EECS Strategic Directions Seminar: The seminar is held on Wednesdays at noon, in 306 Soda Hall. Light refreshments are served. Nov. 9th - Carol Christ Vice Chancellor University of California, Berkeley Nov. 16th - Andy Rappaport Technological Research Group of Boston Nov. 23rd - no seminar Nov. 30th - C.V. Shank Director Lawrence Berkeley Laboratory