Screened Poisson Surface Reconstruction
ACM Transactions on Graphics, 2013 · paper · Computer science
Extends Poisson surface reconstruction, which builds watertight surfaces from oriented points, by adding the points as interpolation constraints through a screened Poisson equation. Algorithmic improvements make the solver's time linear in the number of points, giving faster, higher-quality reconstructions.
Checked: PDF from hhoppe.com, fetched 4 Oct 2026. Page numbers are those of that PDF.
1 finding
Reference or label Section 6.1 says 'to a lesser extent the SSD reconstruction (center left)', but in Fig. 4(a) SSD is the third column (center right); the same sentence also puts the wavelet reconstruction at 'center left'.
- p. 7 Section 6.1, Real Scanner Data, paragraph 2
tent the SSD reconstruction (center left), oversmooth the data, while
- p. 9 Figure 4(a) column labels
Poisson Wavelet SSD Screened
- p. 7 Section 6.1, Real Scanner Data, paragraph on Figure 4(a)
to a lesser ex-tent the SSD reconstruction (center left), oversmooth the data, while the wavelet reconstruction (center left)
- p. 7 Real Scanner Data paragraph
the SSD reconstruction (center left), oversmooth the data, while the wavelet reconstruction (center left)
Judge's note: In Fig. 4(a) on p.6 the columns run Poisson, Wavelet, SSD, Screened. Wavelet is center left and SSD is center right. The text on p.7 calls both 'center left', so the SSD position label is wrong.
Second check: another model agreed
Report a problem with this finding
12 minor (typos and the like; most checked by the judge only)
Reference or label Text gives 'the David (11M points)' (intro: 'a subset of 11.4M points'), but the Fig. 4(b) David panel is labelled 5690141, while the other panels' labels match the text totals.
- p. 7 Section 6.1, Real Scanner Data
the David (11M points)
- p. 6 Figure 4(b), David (RMS) panel, x-axis label
5690141
- p. 1 Introduction
a subset of 11.4M points
Only the authors can say what is right here.
Judge's note: On the p.6 image the labels are Awakening 10292795 (text 10M), Bunny 181008 (0.2M), Lucy 967914 (1.0M) and Neptune 2355718 (2.4M). David is 5690141, about half of 11.4M, so the label convention differs or one of the two values is wrong. The paper does not explain this. Only the authors can say which is meant.
Second check: another model agreed
Report a problem with this finding
Wording Section 6.2: 'For the original and screened Poisson reconstructions, the multigrid solver performs a constant number of conjugate-gradient iterations at each level, giving linear complexity in the number N of octree nodes', while Section 5.1 and 5.3 say the basic/original solver is O(|P|·log|P|) and only the conforming-octree version is linear.
- p. 9 Section 6.2, Running Time
For the original and screened Poisson reconstructions, the multigrid solver performs a constant number of conjugate-gradient iterations at each level, giving linear complexity
- p. 5 Section 5.1
the overall time complexity remains O(|P| · log |P|).
Only the authors can say what is right here.
Judge's note: p.5 (5.1 and 5.3) says the basic algorithm is log-linear because of the coarser-constraint subtraction, and that the conforming octree makes it linear. The intro says the same. Section 6.2 on p.9 credits the original Poisson reconstruction with linear complexity. The relaxation step alone is linear, but the sentence attributes linear complexity to the whole method, which conflicts with Section 5.
Second check: another model agreed
Report a problem with this finding
Wording Text says screened Poisson 'generates reconstruction whose RMS errors are comparable to or smaller than those of the SSD reconstruction', but in the Fig. 4(b) Neptune depth-10 panel the SSD bar is visibly lower (about half) than the Screened bar.
- p. 7 Section 6.1, Real Scanner Data, paragraph on Figure 4(b)
reconstruction whose RMS errors are comparable to or smaller than those of the SSD reconstruction (green)
- p. 6 Fig. 4(b), Neptune (RMS) panel, depth 10
Neptune (RMS)
Judge's note: On the p.6 image, at Neptune depth 10 the SSD (green) bar is about 0.0002 and the Screened (blue) bar about 0.00045, on a 0.002 axis. At Neptune depth 9 they are about equal. Elsewhere Screened is equal or lower. 'Comparable' is loose wording, but it understates this case. Bar heights were read from a low-resolution image.
Second check: another model agreed
Report a problem with this finding
Wording 'When setting α = 0, the new algorithm (shown in brackets) is 2–3 times faster than the original Poisson reconstruction algorithm' does not fit Table I: the ratios run from 0.77 to 3.28, and the bracketed time is slower at depth 8 (10 [13], 41 [45]).
- p. 9 Section 6.2, Running Time, right column
(shown in brackets) is 2–3 times faster than the original Poisson re-
- p. 9 Table I, Poisson time column, depth 8
10 [13]
- p. 9 Table I, Poisson time column, David depth 8
41 [45]
- p. 9 Section 6.2, Running Time, right column
(shown in brackets) is 2–3 times faster than the original Poisson re-construction algorithm
Only the authors can say what is right here.
Judge's note: Read Table I on p.9. Poisson/bracket ratios are about 2–3 only at depth 10, and at depth 11 for Neptune. The depth-8 rows show a slowdown, depth 9 is about 1.5–1.6, and David depth 11 is about 3.3. The claim is unqualified. The point about screened times being faster is handled separately in C009.
Second check: another model agreed
Report a problem with this finding
Wording 'Even with the additional overhead of screening ... our running times are faster than the original Poisson reconstruction' is unqualified, but at depth 8 in Table I Screened is slower (Neptune 14 vs 10 s; David 48 vs 41 s).
- p. 9 Section 6.2, Running Time, last paragraph
running times are faster than the original Poisson reconstruction.
- p. 9 Table I, Neptune and David depth 8 rows
14
- p. 9 Section 6.2, Running Time, last paragraph
our running times are faster than the original Poisson reconstruction.
- p. 9 Table I, depth 8 rows, Time in seconds
41 [45]
Only the authors can say what is right here.
Judge's note: Table I on p.9: Screened beats Poisson only at depths 9–11. The sentence has no depth qualifier. There is no basis on the page for a reviewer's suggestion that 45 is a misprint.
Second check: another model agreed
Report a problem with this finding
Typo The Fig. 9 caption says 'the smoothed signed distance reconstruction', but elsewhere (Section 6, Conclusion, reference title) it is 'smooth signed distance'.
- p. 11 Figure 9 caption
smoothed signed distance reconstruction
- p. 6 Section 6, first paragraph
smooth signed distance reconstruction of Calakli and Taubin
- p. 11 Fig. 9 caption
using the smoothed signed distance reconstruction [Calakli and Taubin 2011]
- p. 6 Section 6, Results
smooth signed distance reconstruction of Calakli and Taubin [2011]
Judge's note: Checked the p.11 caption and p.6 text. The reference list entry reads 'SSD: Smooth signed distance surface...'. This is a cosmetic naming inconsistency.
Second check: none (judge only)
Report a problem with this finding
Typo 'Figure 4(a) shows reconstructions results' should be 'reconstruction results'.
- p. 7 Section 6.1, Real Scanner Data
Figure 4(a) shows reconstructions results for the Neptune and David models
- p. 7 Section 6.1
shows reconstructions results for the Neptune
Judge's note: Visible on the p.7 page image, Section 6.1, Real Scanner Data, second paragraph.
Second check: none (judge only)
Report a problem with this finding
Typo 'and generates reconstruction whose RMS errors' is missing an article or plural ('a reconstruction' / 'reconstructions').
- p. 7 Section 6.1
generates reconstruction whose RMS errors
Judge's note: Visible on p.7, right column, first paragraph.
Second check: none (judge only)
Report a problem with this finding
Typo 'the condition number of the resulting system matrix is square that of the Laplacian matrix' should be 'is the square of that of the Laplacian matrix'.
- p. 9 Section 6.2
system matrix is square that of the Laplacian matrix
Judge's note: Visible on p.9, left column, last paragraph.
Second check: none (judge only)
Report a problem with this finding
Typo 'with sufficiently fine decretization' should be 'discretization'.
- p. 10 Section 6.3, Screening Parameter α
fine decretization
- p. 10 Section 6.3, Screening Parameter α
sufficiently fine decretization
- p. 10
with sufficiently fine decretization, both its energy terms
Judge's note: The misspelling is present on the p.10 image (Section 6.3). 'discretization' is used everywhere else in the text.
Second check: none (judge only)
Report a problem with this finding
Typo Kolluri et al. 2004 reference title is printed as "Spectral surface reconstruction from noise point clouds"; "noise" should read "noisy".
- p. 12 References, Kolluri et al. 2004
reconstruction from noise point clouds
- p. 12 References, Kolluri et al. 2004
Spectral surface reconstruction from noise point clouds.
- p. 12 References, Kolluri et al. 2004
Spectral surface reconstruction from noise point clouds
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: the published SGP 2004 title is "Spectral Surface Reconstruction from Noisy Point Clouds". The page 12 image shows "from noise point clouds" as printed, so this is not an extraction artefact. "noise point clouds" is also ungrammatical on its face. Cosmetic typo in the reference title.
Second check: none (judge only)
Report a problem with this finding
Typo Carr et al. 2001 reference: "MCCALLUM, B., and EVANS, T. 2001" sets the conjunction in ordinary roman type with an extra space, while every other entry uses small-caps "AND".
- p. 12 References, Carr et al. 2001
B., and EVANS, T. 2001
Judge's note: In the page 12 image, the "and" in the Carr et al. author list is plain lowercase roman with a double space before it. In every neighbouring entry (Calakli and Taubin, Cazals and Giesen, Chuang and Kazhdan, etc.) the conjunction is in small caps. This is a purely cosmetic formatting inconsistency.
Second check: none (judge only)