Conception et étude d'une architecture parallèle à réseau linéaire de processeurs et mémoires CAM pour le traitement d'image

Abstract : The works on characterization of objects in an image incited the design of a dedicated architecture for connected-component labeling. Due to both local and global properties, this operation is extremely time costly when a sequential approach is applied. In the case of a parallel approach, it has been demonstrated that the labeling complexity is inversely proportional to the number of processors. In order to reduce the complexity, many of the existing solutions require a large number of processors. However, the "Processor-Time" product has to be considered to obtain an optimal architecture particularly if VLSI implementation is envisaged. The primary objective of this work was to design a parallel dedicated architecture for labeling with both weak product and good tradeoff of "Processor-Time". For this, a Maresca algorithm, implemented in a 2-D array of processors, Polymorphic-Torus was adapted. The structure obtained has a 1-D array of processors with CAM memories. [...]
Document type :
Theses
File URL :
http://docnum.univ-lorraine.fr/prive/SCD_T_1997_0116_MOZEF.pdf
Complete list of metadatas

https://hal.univ-lorraine.fr/tel-01747348
Contributor : Thèses Ul <>
Submitted on : Thursday, March 29, 2018 - 10:57:43 AM
Last modification on : Tuesday, March 12, 2019 - 3:52:04 PM

Identifiers

  • HAL Id : tel-01747348, version 1

Collections

Citation

Eril Mozef. Conception et étude d'une architecture parallèle à réseau linéaire de processeurs et mémoires CAM pour le traitement d'image. Autre [cond-mat.other]. Université Henri Poincaré - Nancy 1, 1997. Français. ⟨NNT : 1997NAN10116⟩. ⟨tel-01747348⟩

Share

Metrics

Record views

1