VLSI implementation of anisotropic probabilistic neural network for real-time image scaling

Ching Han Chen, Hsiang Wen Chang, Chia Ming Kuo

Research output: Contribution to journalArticlepeer-review

3 Scopus citations

Abstract

This study proposes an VLSI implementation of anisotropic probabilistic neural network (APNN) for real-time video processing applications. The APNN interpolation method achieves good sharpness enhancement at edge regions and reveals the noise reduction at smooth region. For real-time applications, the APNN interpolation is further implemented with efficient pipelined very-large-scale integration (VLSI) architecture. The VLSI architecture of APNN has a five-layer structure, which is comprised of Euclidian layer, Gaussian layer, weighting layer, summation layer, and division layer. The VLSI implementation outperforms software with the low-loss quality. The experimental results indicate that the performance of VLSI implementation is competent for image interpolation. The presented VLSI implementation of APNN interpolation method can reach 1920 × 1080 at 30 frames per second (FPS) with a reasonable hardware cost.

Original languageEnglish
Pages (from-to)71-80
Number of pages10
JournalJournal of Real-Time Image Processing
Volume16
Issue number1
DOIs
StatePublished - 14 Feb 2019

Keywords

  • Anisotropic
  • Interpolation
  • Neural networks
  • Sharpness
  • Smoothness

Fingerprint

Dive into the research topics of 'VLSI implementation of anisotropic probabilistic neural network for real-time image scaling'. Together they form a unique fingerprint.

Cite this