No-reference video quality assessment design framework based on modular neural networks

  • Authors:

  • Dragan D. Kukolj;Maja Pokrić;Vladimir M. Zlokolica;Jovana Filipović;Miodrag Temerinac

  • Affiliations:

  • Faculty of Technical Sciences, University of Novi Sad, Novi Sad, Serbia;Faculty of Technical Sciences, University of Novi Sad, Novi Sad, Serbia;Faculty of Technical Sciences, University of Novi Sad, Novi Sad, Serbia;Faculty of Technical Sciences, University of Novi Sad, Novi Sad, Serbia;Faculty of Technical Sciences, University of Novi Sad, Novi Sad, Serbia

  • Venue:
  • ICANN'10 Proceedings of the 20th international conference on Artificial neural networks: Part I
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper presents a novel no-reference video quality assessment (VQA) model which is based on non-linear statistical modeling. In devised nonlinear VQA model, an ensemble of neural networks is introduced, where each neural network is allocated to the specific group of video content and features based on artifacts. The algorithm is specifically trained to enable adaptability to video content by taking into account the visual perception and the most representative set of objective measures. The model verification and the performance testing is done on various MPEG-2 video coded sequences in SD format at different bit-rates taking into account different artifacts. The results demonstrate performance improvements in comparison to the state-of-the-art nonreference video quality assessment in terms of the statistical measures.