Item type |
SIG Technical Reports(1) |
公開日 |
2022-01-20 |
タイトル |
|
|
タイトル |
Bilateral Normal Integration |
タイトル |
|
|
言語 |
en |
|
タイトル |
Bilateral Normal Integration |
言語 |
|
|
言語 |
eng |
キーワード |
|
|
主題Scheme |
Other |
|
主題 |
画像処理・機械学習 |
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 |
|
|
|
Osaka University |
著者所属 |
|
|
|
Osaka University |
著者所属 |
|
|
|
Peking University |
著者所属 |
|
|
|
Osaka University |
著者所属(英) |
|
|
|
en |
|
|
Osaka University |
著者所属(英) |
|
|
|
en |
|
|
Osaka University |
著者所属(英) |
|
|
|
en |
|
|
Peking University |
著者所属(英) |
|
|
|
en |
|
|
Osaka University |
著者名 |
Xu, Cao
Hiroaki, Santo
Boxin, Shi
Yasuyuki, Matsushita
|
著者名(英) |
Xu, Cao
Hiroaki, Santo
Boxin, Shi
Yasuyuki, Matsushita
|
論文抄録 |
|
|
内容記述タイプ |
Other |
|
内容記述 |
This paper studies the discontinuity preservation problem in the task of surface recovery by integrating a single-view surface normal map. Although various sophisticated strategies have been proposed, this challenging problem has not been well solved due to unknown discontinuity positions in the normal map. The key of our method is to model the existence of discontinuity between every two adjacent pixels by the depth differences between that two pixels. At each pixel, we approximate the left and right (resp., upper and lower) partial derivatives and relatively weight the two approximations by comparing the depth differences at the left and right (resp., upper and lower) sides. Therefore we term our method as “bilateral” normal integration. By iteratively solving for the depths and updating the discontinuity maps, we can recover the surface with discontinuities. Experiments on various challenging surfaces demonstrate our method's effectiveness. In addition, we show that discontinuities can be preserved in perspective normal maps by our method. |
論文抄録(英) |
|
|
内容記述タイプ |
Other |
|
内容記述 |
This paper studies the discontinuity preservation problem in the task of surface recovery by integrating a single-view surface normal map. Although various sophisticated strategies have been proposed, this challenging problem has not been well solved due to unknown discontinuity positions in the normal map. The key of our method is to model the existence of discontinuity between every two adjacent pixels by the depth differences between that two pixels. At each pixel, we approximate the left and right (resp., upper and lower) partial derivatives and relatively weight the two approximations by comparing the depth differences at the left and right (resp., upper and lower) sides. Therefore we term our method as “bilateral” normal integration. By iteratively solving for the depths and updating the discontinuity maps, we can recover the surface with discontinuities. Experiments on various challenging surfaces demonstrate our method's effectiveness. In addition, we show that discontinuities can be preserved in perspective normal maps by our method. |
書誌レコードID |
|
|
収録物識別子タイプ |
NCID |
|
収録物識別子 |
AA11131797 |
書誌情報 |
研究報告コンピュータビジョンとイメージメディア(CVIM)
巻 2022-CVIM-228,
号 17,
p. 1-8,
発行日 2022-01-20
|
ISSN |
|
|
収録物識別子タイプ |
ISSN |
|
収録物識別子 |
2188-8701 |
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 |