Efficient Video Coding Based on Projective Transform and Object Tracking

DOI

Bibliographic Information

Other Title
  • 射影変換と物体追跡に基づく効率的映像符号化

Abstract

Even uncompressed, original video signals contain camera noises and such noise works hard on coding efficiency. We propose to track rigid objects in video with a simple motion model and temporally align them to reduce camera noises and to increase coding efficiency. In our experiments, this preprocessing worked good for both objective and subjective metrics. In terms of luminance BD-Rate, our method gained 31-41% against HEVC reference software HM16.4 and 26-33% against JVET (future video coding development team) reference software JEM2.0. In terms of subjective video quality, decoded video looked even better than original video with only using 140-240 times less bits than HM16.4 bitstreams.

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top