| Item type |
SIG Technical Reports(1) |
| 公開日 |
2023-08-31 |
| タイトル |
|
|
タイトル |
動的メッシュ圧縮におけるLoDベース変位量並び替えに関する検討 |
| タイトル |
|
|
言語 |
en |
|
タイトル |
A Study on LoD-based Reordering of Displacements Coefficients in Dynamic Mesh Compression |
| 言語 |
|
|
言語 |
jpn |
| 資源タイプ |
|
|
資源タイプ識別子 |
http://purl.org/coar/resource_type/c_18gh |
|
資源タイプ |
technical report |
| 著者所属 |
|
|
|
KDDI総合研究所 |
| 著者所属 |
|
|
|
KDDI総合研究所 |
| 著者所属 |
|
|
|
KDDI総合研究所 |
| 著者所属(英) |
|
|
|
en |
|
|
In the video-based dynamic mesh coding, the mesh is divided into a base mesh and detailed displacements. The level values of displacement component coefficients are packed into an image and then images are encoded with an existing video codec. Conventional methods for image packing have changed the starting position of packing from the top-left to the bottom-right of the image. This change ensured that level values in lower LoDs, which tend to have large amplitude values, are positioned in the bottom-right of the image, leading to improved coding efficiency. However, the amplitude of the level values of LoD=1 often becomes larger than those of the lower frequency LoD=0 or those of higher frequency LoD=2 and above. Consequently, the boundaries between the LoD=1 and other LoDs become more noticeable, and it is difficult to predict the LoD=0 level values from LoD1, making the image difficult to compress. Therefore, this paper proposes to reorder the level values of LoD0 and LoD1 for all frames or adaptive frames to make the packed image more compressible. Experimental results on MPEG test sequences confirmed that a maximum gain was -0.043% compared to the conventional image packing method. |
| 著者名 |
明堂, 絵美
西村, 仁志
河村, 圭
|
| 著者名(英) |
Emi, Myoucou
Hitoshi, Nishimura
Kei, Kawamura
|
| 論文抄録 |
|
|
内容記述タイプ |
Other |
|
内容記述 |
動的メッシュ圧縮方式では，メッシュは基本メッシュと詳細な変位量に分割される．変位量の係数レベル値は，画像にパッキングされた後に既存のビデオコーデックで符号化される．従来の変位量係数レベル値の画像パッキング手法では，パッキング開始箇所を，画像の左上から右下に変更した．これにより，詳細度 Level of Detail (LoD) が小さく，大きな振幅値となりやすい変位量係数レベル値が画像の右下に配置されやすくなり，符号化効率が向上した．しかし，LoD＝1 の変位量係数レベル値の振幅は，それより低周波の LoD＝0 やそれより高周波の LoD＝2 以上の変位量係数レベル値の振幅より大きくなりやすい．そのため，LoD＝1 と他の LoD の変位量係数レベル値の境界が目立ちやすく，LoD＝1 から LoD＝0 の変位量係数レベル値を予測しにくいなど，圧縮しにくい画像となっていた．そこで，本稿では，全フレームまたは適応的に LoD＝0 と LoD＝1 の変位量係数レベル値の配置を並び替え，圧縮しやすい画像とする手法を提案する．実験では，MPEG テストシーケンスにおいて，従来の画像パッキング手法と比較して最大 -0.043% の符号化効率の向上を確認した． |
| 論文抄録(英) |
|
|
内容記述タイプ |
Other |
|
内容記述 |
In the video-based dynamic mesh coding, the mesh is divided into a base mesh and detailed displacements. The level values of displacement component coefficients are packed into an image and then images are encoded with an existing video codec. Conventional methods for image packing have changed the starting position of packing from the top-left to the bottom-right of the image. This change ensured that level values in lower LoDs, which tend to have large amplitude values, are positioned in the bottom-right of the image, leading to improved coding efficiency. However, the amplitude of the level values of LoD=1 often becomes larger than those of the lower frequency LoD=0 or those of higher frequency LoD=2 and above. Consequently, the boundaries between the LoD=1 and other LoDs become more noticeable, and it is difficult to predict the LoD=0 level values from LoD1, making the image difficult to compress. Therefore, this paper proposes to reorder the level values of LoD0 and LoD1 for all frames or adaptive frames to make the packed image more compressible. Experimental results on MPEG test sequences confirmed that a maximum gain was -0.043% compared to the conventional image packing method. |
| 書誌レコードID |
|
|
収録物識別子タイプ |
NCID |
|
収録物識別子 |
AN10438399 |
| 書誌情報 |
研究報告オーディオビジュアル複合情報処理（AVM）
巻 2023-AVM-122,
号 3,
p. 1-4,
発行日 2023-08-31
|
| ISSN |
|
|
収録物識別子タイプ |
ISSN |
|
収録物識別子 |
2188-8582 |
| Notice |
|
|
|
SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc. |
| 出版者 |
|
|
言語 |
ja |
|
出版者 |
情報処理学会 |
IPSJ-AVM23122003.pdf (1.3 MB)
