Neuron-based VLSI architecture for real-time camera distortion correction

C. H. Chen, T. K. Yao

Research output: Contribution to journalArticlepeer-review

Abstract

This paper proposes the efficient VLSI architecture of camera distortion correction, based on a Neural Camera Distortion Model (NCDM). Conventional imaging methods use over two kinds of models to correct the camera and lens distortions, but the NCDM uses a single model to immediately correct the geometry distortion and unsymmetrical manufacturing errors. The NCDM, with four neurons, performs a wideangle distortion correction. The results show that the maximal corrected error in a whole image is less than 1.1705 pixels, and the MSE approaches 0.1743 between corrected and ideal results. The distortion correction by NCDM is 429x more accurate than the conventional approach. The chip size of NCDM is 1:51 x 1:51 mm2 and contains 126 K gates using the TSMC 90 nm CMOS technology process. Working at 240 Mhz, this architecture can correct 30 frames and a Full-HD resolution video per second. Results show that the maximal corrected error in a whole image is less than 1.4 pixels, and the mean square error approaches 0.0376 between corrected and ideal results.

Original languageEnglish
Pages (from-to)2150-2162
Number of pages13
JournalScientia Iranica
Volume22
Issue number6
StatePublished - 2015

Keywords

  • Distortion correction
  • Neural network
  • VLSI
  • Wide-angle view

Fingerprint

Dive into the research topics of 'Neuron-based VLSI architecture for real-time camera distortion correction'. Together they form a unique fingerprint.

Cite this