No Reference Video Quality Assessment Based on Least Squares Support Vector Machines

EasyChair Preprint 9451

Abstract

Now a days, need of Application developers towards developing front end-based video applications like Skype or others which forced in huge competition between quality of service and experience. Out of all existing approaches, we considered no reference video quality assessment and moreover, our interest lies towards formulating and melding effective features into one model based on human visualizing characteristics. This research explores the tradeoffs between quality prediction and complexity towards identifying sparseness of LSSVM model and also involves in feature extraction of h.264-bit stream information extracted at macro block layer towards building up of a machine learning based model for quality Assessment. These features which are expected to have high correlation with the perceptual quality of the videos and We concluded that our proposed model outperformed in terms of performance but only in the case of subjective quality assessment and more over due to refining process of subjective scores, fault in encoding process was traced out which is based on error concealment and in case of building up of proposed model with SSIM and MS-SSIM metric at frame level sparseness was traced out.

Keyphrases: LS-SVM, MS-SSIM, NR-VQM, SSIM, SVM, VQM

@booklet{EasyChair:9451,
  author    = {Amitesh Kumar Singam and Venkat Raj Reddy Pashike},
  title     = {No Reference Video Quality Assessment Based on Least Squares Support Vector Machines},
  howpublished = {EasyChair Preprint 9451},
  year      = {EasyChair, 2022}}