Implementation of Median Filters

Chair: Edward J. Coyle, Purdue University, USA

Home

A Continuous-Time Analog Median Filter

Authors:

Alejandro Diaz-Sanchez, Centro Nacional de Investigacion y Desarrollo Tecnologico (Mexico)
Jaime Ramirez-Angulo, New Mexico State University (U.S.A.)
Antonio Lopez, Universidad Publica de Navarra (Spain)

Volume 1, Page (NA), Paper number 641

Abstract:

A continuous-time analog median filter implementation is presented. The filter uses two new circuits to implement the delay line and median blocks. Both circuits were designed in a compact and modular structure to allow implementation of array processors. The circuit was designed for a 2um technology MOSIS process.

ns970641.pdf (Scanned)

TOP

VLSI Cellular Neural Network Based Median Filter for Real-Time Video Signal Processing

Authors:

Jacek Kowalski, Technical University of Lodz (Poland)
Krzysztof Slot, Technical University of Lodz (Poland)
Tomasz Kacprzak, Technical University of Lodz (Poland)

Volume 1, Page (NA), Paper number 642

Abstract:

An increasing demands for video application in teleconferencing, multimedia services, autonomous navigation, motion analysis, object tracking, surveillance, astronomical, and medical imaging make high rigorous requirements for real-time noise filtering algorithms to improve the visual quality and the performance of subsequent processing tasks such as coding, analysis and interpretation. Median based filters - as it was well reported in literature - belong to the class of nonlinear filters and are very useful for these applications. This is because of two intrinsic properties of median filters: edge preservation while reducing efficiently the noise. Even though the theory and computer simulation of median filters are widely studied in the literature [1], the hardware implementation problem still presents opened questions. In VLSI CMOS implementation two concepts are considered: digital [2] and analog [3]. Digital approach (using DSP or FPGA) is very accurate, however is time-consuming and occupies large silicon area which makes this concept expensive for today needs of real-time applications. In comparison with this approach an analog version seems to be much more economic, and even though it is not so accurate as digital one, it can be accepted for many applications. Therefore, the problem of analog realizations of median filters is worth to study in details.