Video Coding using Multiple Reference Frames and Weighted Averaging


Creative Commons License

Kilinç Ç., Seke E.

2023 Innovations in Intelligent Systems and Applications Conference (ASYU), Sivas, Türkiye, 11 - 13 Ekim 2023, ss.1-4

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/asyu58738.2023.10296633
  • Basıldığı Şehir: Sivas
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.1-4
  • Eskişehir Osmangazi Üniversitesi Adresli: Evet

Özet

In contemporary video coding methods, an important step is to find difference between the current data element with a previously coded-decoded data element. This is achieved by referencing previously encoded similar blocks to minimize the data required to represent the new block. The development of the High Efficiency Video Coding (HEVC) standard drew upon research conducted on Advanced Video Coding (AVC), a modern video coding approach. While HEVC and AVC share similarities, HEVC surpasses AVC in terms of digital video encoding capabilities. In this study, improvements were made to the inter-frame structures by utilizing the strategy of HEVC's multiple reference frames. The proposed approach utilized Sum of Absolute Differences (SAD) to select up to three reference blocks. These references are then combined using weighted average to generate an artificial reference for the block to be encoded. The method of least squares was employed to calculate the weights of the selected reference blocks. Experimental results demonstrated that increasing the number of reference blocks up to three resulted in an improvement in the signal-to-noise ratio (PSNR). Consequently, achieving the same PSNR required less data due to this enhancement.