A New Data Format for Multiview Video

• MehrdadPanahpourTehrani

• Akio Ishikawa

• Masashiro Kawakita

• Naomi Inoue

• Toshiaki Fujii

抄録

This paper proposes a new data forma that can be used for multiview video representation. The new data format compensates the synthesis error at the viewpoint where the image is converted to a new format. In this scenario, residual based representations have already been proposed. In this paper we introduce and further discuss the new data format that we have recently proposed in our previous work. The converted image to the new data format is called hybrid image. NICT has developed a 200-inch 3D display. This display requires 200 HD multi-view images per frame at display side. We choose the hybrid representation as a novel data format for transmission of multi-view images to the display. A hybrid image consists of residual and reminder values, while the state-of-art use either residual or reminder. The representation based on hybrid image is a sequence of original and hybrid images that are repeated consequently in view/time direction. We evaluate the compression efficiency of represented multi-view images by hybrid data format. In this paper, we discuss a suitable camera configuration for our 3D display. Furthermore, we compare the compression efficiency of hybrid representation against other representations. Experiments demonstrate improvement of coding efficiency about 20% in average, using hybrid representation.This paper proposes a new data forma that can be used for multiview video representation. The new data format compensates the synthesis error at the viewpoint where the image is converted to a new format. In this scenario, residual based representations have already been proposed. In this paper we introduce and further discuss the new data format that we have recently proposed in our previous work. The converted image to the new data format is called hybrid image. NICT has developed a 200-inch 3D display. This display requires 200 HD multi-view images per frame at display side. We choose the hybrid representation as a novel data format for transmission of multi-view images to the display. A hybrid image consists of residual and reminder values, while the state-of-art use either residual or reminder. The representation based on hybrid image is a sequence of original and hybrid images that are repeated consequently in view/time direction. We evaluate the compression efficiency of represented multi-view images by hybrid data format. In this paper, we discuss a suitable camera configuration for our 3D display. Furthermore, we compare the compression efficiency of hybrid representation against other representations. Experiments demonstrate improvement of coding efficiency about 20% in average, using hybrid representation.

収録刊行物

• 研究報告オーディオビジュアル複合情報処理（AVM）

  研究報告オーディオビジュアル複合情報処理（AVM） 2013 (6), 1-5, 2013-02-15

詳細情報

• CRID: 1572261552818004096
• NII論文ID: 110009536462
• NII書誌ID: AN10438399
• 本文言語コード: en
• データソース種別:
  • CiNii Articles