Output details
13 - Electrical and Electronic Engineering, Metallurgy and Materials
University of Sheffield : A - Electronic and Electrical Engineering
An Overview and Performance Evaluation of Classification-Based Least Squares Trained Filters
Algorithm proposed for optimising filter coefficients based on an off-line training procedure and dedicated image content classification for various image restoration/enhancement applications, such as denoising, sharpening and resolution enhancement. The off-line training is efficient and the online filtering is real-time. The proposed 'Trained Filters' perform significantly better than many heuristically-designed filters. Algorithm has been implemented in ASIC for Philips high-definition TV sets (IhorKirenko: ihor.kirenko@philips.com) and in FPGA for BOSCH video surveillance cameras (SachaCvetković: sacha.cvetkovic@nl.bosch.com). Two EU patent applications (07300718.9, 06123138.7) have been filed. This also led to a PhD studentship funded by Bosch (JanKlijn: jan.klijn@nl.bosch.com).