VMAF is a perceptual video quality assessment algorithm developed by Netflix. The project aims to create a fully functional video filter inside FFmpeg that measures the Video Multimethod Assessment Fusion (VMAF) score of any input video compared to a reference. The overall goal is to have both the independent vmaf containing the implementation of metrics used in vmaf like ansnr,vif,dlm,motion and fusing them using Support Vector Machine (SVM) and a dependent library implementation linking to Netflix's vmaf (librarification).

Organization

Student

Ashish Pratap Singh

Mentors

  • Ronald S. Bultje
close

2017