Fraunhofer-Gesellschaft

Publica

Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

A framework for the adaptation of image operators

 
: Köppen, M.; Vicente-Garcia, R.

Cagnoni, S.:
Genetic and evolutionary computation for image processing and analysis
New York, NY: Hindawi, 2007 (EURASIP book series on signal processing and communications 8)
ISBN: 977-454-001-8
ISBN: 978-977-454-001-1
pp.215-239
English
Conference Paper
Fraunhofer IPK ()

Abstract
A framework was presented, which allows for the design of texture filters for fault detection (two class problem). The framework is based on the 2D-Lookup algorithm, where two filter output images are used as input. "figure presented" "figure presented" The approach can be applied to a large class of texture analysis problems. The results, obtained without "human intervention," are ready-to-use texture filters. Also, they can be tuned in order to obtain even more better results, or combined in a superposed inspection system. The following are our experiences during the use of the system. (i) The framework was able to design texture filters with good or very good performance. (ii) The goal image matched the fault region quite satisfactorily. (iii) Bordering regions should be neglected for fitness evaluation. (iv) The framework was able to design filters for the detection of noncompact fault regions and fault regions with varying appearance. (v) The designed filters may b e subjected to further improvements by the user. "figure presented" Improvements of the whole architecture were considered as well: one is based on an evaluation of the 2D-Lookup matrix by neural networks in order to get a more comprehensive solution for a given texture filtering problem, the other for extending the application scope to low-contrast texture fault processing, that is, faults which are hard to separate from the background texture. The second extension of the framework is a two-stage one, based on 2D histogram lookup and consecuting 2D-Lookup adaptation.

: http://publica.fraunhofer.de/documents/N-171569.html