Nonlinear relaxation networks for preprocessing and segmenting

Project Supervisor: Prof. Dr.-Ing. R.-R. Grigat
Coworker: Dr. Dipl.-Ing. K. Wiehler
Running time: 7.1996 - 12.2000
Financing: DFG
Publications: here
References: CNN, nonlinear image processing, nonlinear diffusion, Analog VLSI

In a system to the picture preprocessing intoxication reduction and high information flow-rate are important requirements. By analog Design of cellular neural one nets these requirements with working times within the microsecond range could be realisierzt. The algorithm is based on nonlinear diffusion and suppresses noise, while signal slopes remain.

A Multi FPGA system for the evaluation of image processing algorithms

Project Supervisor: Prof. Dr.-Ing. R.-R. Grigat
Coworker: Dipl.-Ing. M. Perezowsky
Running time: 1.1997 - 12.2000
Financing: TUHH
Publications: here
References: Multi-FPGA-System, System verification, Substantial parallel image processing

In the context of this study develops a reconfigurable hardware system, which offers the possibility of evaluating under real time requirements complex image processing algorithms. Programming takes place in VHDL, so that the verified systems are suitable as basis of a VLSI Design. The hardware system is based on the PCI bus, which heads for up to four modular arithmetic and logic units with four FPGAs each. The datentransfer is decoupled and by a fifth FPGA is steered with a bi-directional Fifo of the PCI bus. Also different arithmetic and logic units (e.g. Risc CCU) can be used by the modular structure in the system. The four arithmetic and logic units are equipped with local memory and cyclically with one another connected. Thus the system is particularly suitable for the substantial parallel image processing.

Color Management

Project Supervisor: Prof. Dr.-Ing. R.-R. Grigat
Coworker: Dr. Dipl.-Math. Philip Urban
Running time: since 11.1999
Financing: Industrie
Publications: here
References: Color area, scanner, monitor, color printer, color perception

The project is concerned with the Reporduktion of colored collecting mains. The achievement of the same color impression is located independently of monitor, type of printer, type of paper and type of scanner in the center of the investigations. Beyond that the reproduction of colored collecting mains in the black-and-white pressure is optimized.

Fast image processing of medical x-ray image sequences

Project Supervisor: Prof. Dr.-Ing. R.-R. Grigat
Coworker: Dr. Dipl.-Ing. Marc Hensel, Dipl.-Ing. André Gooßen
Running time: 04.2003 - 03.2007, 12.2007 - 11.2010
Financing: Industrie
Publications: M. Hensel, A. Gooßen
References: x-ray imaging, radiographs, real-time processing, image sequences, multi-processor systems, noise reduction

Fast image processing of medical x-ray image sequences Generating subjectively "good" radiographs requires enhancement of structures and noise reduction.

In principle, noise can be reduced by applying a high x-ray dose rate. However, in medical applications the rate is to be kept as small as possible in order to avoid excessive exposure of the patients. The problem matters in particular in dynamic radiographs, since the patients are exposed to the radiation during a longer period. Therefore comparative weak dose rates are used, whereby the images contain strong noise.

The research goal is real time noise reduction in radiograph sequences under utilization of

spatial
temporal
band-dependent

image properties.

Algorithms for Digital Cameras in Mobile Phones

Project Supervisor: Prof. Dr.-Ing. R.-R. Grigat
Coworker: Dipl.-Ing. Touraj Tajbakhsh
Running time: 07.2005 - 07.2008
Financing: Industrie
Publications: here
References: digital image processing, color management, image stabilisation, auto-exposure-control, autofocus

Digital cameras are increasingly deployed in mobile phones and PDAs. The optics and optical sensors used in these devices are small and thus achieve a poor SNR and create severe smear artifacts by small movements. The often utilized inexpensive lens systems cause geometrical distortions and chromatic aberrations. The objective of this project is the simulation of a complete ISP chain with MATLAB Simulink to improve image quality. New control loops and algorithms in image processing will be developed and verified to have a reference for a VHDL implementation.

Learning an invariant distance metric

Visiting researcher: M.Sc. Zahra Goudarzi
Running time: 02.2017 - 31.12.2018
Financing:
Publications: hier
References: Metric learning, Mahalanobis Metric, Grassmann manifolds

The requirement for suitable ways to measure the distance or similarity between data is omnipresent in machine learning, pattern recognition and data mining, but extracting such good metrics for particular problems is in general challenging.

This has led to the emergence of metric learning ideas, which intend to automatically learn a distance function tuned to a specific task. In many tasks and data types, there are natural transformations to which the classification result should be invariant or insensitive. This demand and its implications are essential in many machine learning applications, and insensitivity to image transformations was in the first place achieved by using invariant feature vectors.

Aim of this project is to learn a metric which is invariant to the different transformations such as horizontal translation, vertical translation, global scale, rotation, line thickness, and shear and also illumination changes that might be applied on data. To do so, the first idea is taking the advantage of the Projection metric on the Grassmann manifolds.