Programs in Physics & Physical Chemistry
|[Licence| Download | New Version Template] acbj_v1_0.gz(31 Kbytes)|
|Manuscript Title: A multiresolution noise removal algorithm for visual pattern recognition in imaging detectors.|
|Authors: M. Castellano, E. Nappi, F. Posa, G. Tomasicchio|
|Program title: PRIP_ENHANCE|
|Catalogue identifier: ACBJ_v1_0|
Distribution format: gz
|Journal reference: Comput. Phys. Commun. 66(1991)293|
|Programming language: C, Fortran.|
|Computer: VAX 6340.|
|Operating system: VMS VERSION 5.2.|
|RAM: 515K words|
|Word size: 32|
|Keywords: Particle physics, Elementary, Event reconstruction, Filtering, Image enhancement, Spot detection, Visual pattern Recognition.|
Nature of problem:
High multiplicity events, generated by recent physics experiments require high granularity detectors in particular those read out by CCD's. These devices produce unambiguous bidimensional representations of spot-composed events thus involving the use of digital image process- ing techniques for data analysis. Random noise from light reflections and CCD noise arise in such a way to elude traditional noise removal methods making unreliable further event pattern recognition strategies.
A filtering-based preprocessing algorithm to discern selectively noise from relevant gray level spot structures of images, produced by high granularity detectors in HEP experiments, is proposed. It carries out a fine multiresolution survey of image through a multistep approach based on the Hierarchical Discrete Correlation (HDC) technique. A set of enhanced images generated by the original ones is produced and made available for further pattern recognition phases.
Raw images show a highly noisy background from which emerges, by a visual inspection, the expected pattern scattered according to intensity and spatial relationships among the items composing it. The best working conditions for the program are determined by the use of spot- composed images usually corrupted by fine-grained or more isolated structures of noise.
The running time depends on the size of the input image. The application of the algorithm on a 208 X 144 pixels image gives rise to the following results:
tFL(i) ~6i s; tEN(i) ~i(3i+5) s, (i=0,1...10);where tFL(i) and tEN(i) are the times requested at i-th level of the filtering and enhancement respectively.
|Disclaimer | ScienceDirect | CPC Journal | CPC | QUB|