1. What have the authors contributed in "Error analysis for image-based rendering with depth information" ?
The authors quantitatively analyze the rendering quality of image-based rendering ( IBR ) algorithms using per-pixel depth.. Assuming the ideal pinhole camera model, the authors show that IBR errors can be quantified using IBR configurations such as the depth and intensity estimate errors, the scene geometry and texture, the number of actual cameras, their positions and resolution.. The authors discuss the implications of the proposed analysis on camera placement, budget allocation, and bit allocation.. In particular, the proposed analysis suggests that, in smooth regions, the decay rates of the virtual images ’ mean absolute errors are O ( λ ) and O ( λ ), where λ is the local density of actual samples, for 2D and 3D scenes, respectively.
read more
2. What have the authors stated for future works in "Error analysis for image-based rendering with depth information" ?
Finally, to extend Lemma 5 in the presence of sample errors and jitters, it is sufficient to prove the following lemma.
read more
3. What is the L norm for the virtual image?
since the virtual image is discontinuous, the authors use the 95%-point value (instead of the 100%-point value, i.e. the maximum) of the convolution as the L∞ norm.
read more
4. how can the authors compute the moment E[Sk]?
The circumradius R has the probability density function (pdf)2(πλ)2r3e−πλr 2 , r > 0. (53)The moments E[Sk] can be computed using explicit formula.
read more
![Fig. 10. The ground truth image of the scene at the virtual camer Cv = [4, 0, 0]T .](/figures/fig-10-the-ground-truth-image-of-the-scene-at-the-virtual-i39vshjd.webp)
![Fig. 1. The calibrated scene-camera model. The scene surface is modeled as a 2D parameterized curveS(u) (Fig. 1(a)) or as a 3D parameterized surface S(u, v) (Fig. 1(b)). The texture map, denotedT (u) for the 2D case orT (u, v) for the 3D case, is “painted” on the object surface. We assumethe pinhole camera model with calibrated positional matrixΠ = [R,T ] ∈ R2×3 or Π ∈ R3×4 for the 2D case or the 3D case, respectively. Finally, the camer resolution is characterized by the pixel interval∆x on the image line (2D case), or by the pixel intervals∆x,∆y in horizontal and vertical direction on the image plane (3D case).](/figures/fig-1-the-calibrated-scene-camera-model-the-scene-surface-is-3viqeopd.webp)
