Adaptive Multiplane Image Generation from a Single Internet Picture

Luvizon, Diogo C.; Carvalho, Gustavo Sutter P.; Santos, Andreza A. dos; Conceição, Jhonatas S.; Flores-Campana, Jose L.; Decker, Luis G. L.; Souza, Marcos Roberto e; Pedrini, Hélio; Joia, Antonio; Penatti, Otávio Augusto Bizetto

doi:10.1109/wacv48630.2021.00260

Cited by 16 publications

(4 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The textured proxies are warped and blended from back to front using alpha compositing with the "over" operator so that the stacked proxies represent full parallax under 6 degrees of freedom camera motion. Given differently shaped proxies that sweep the volumetric space, variants can infer MPI or MSI from a perspective [31,54], stereo [1,50,60], and multi-view images [13,33,51]. Combined with more explicit depth geometry prior, we could expect improved rendering quality with additional costs [10,21,28,61].…”

Section: Multi-layer Scene Representationmentioning

confidence: 99%

Multi-Layer Scene Representation from Composed Focal Stacks

Ishikawa,

Saito,

Kalkofen

et al. 2023

IEEE Trans. Visual. Comput. Graphics

View full text Add to dashboard Cite

Fig. 1: We propose representing scenes as composed focal stacks, which we compute from registered images, for view synthesis. Our approach enables diminishing local artifacts such as motion blur and ISO noise, to improve multi-layer scene representations. (a) Our approach allows the user to randomly capture photos in a continuous motion. (b) From the captured images, we compose a synthetic focal stack, from which we derive a multi-layer scene representation using a CNN. (c) The outcome of the network is a multi-layer scene representation, which enables continuous viewpoint changes. (d) Our approach supports several applications, including 6 degrees of freedom wide field of view scene representation, which enables photo-realistic VR applications (left), scene representations from underexposed images (right), and multi-layer image generation from captured focal stacks (top).

show abstract

Section: Multi-layer Scene Representationmentioning

confidence: 99%

Multi-Layer Scene Representation from Composed Focal Stacks

Ishikawa,

Saito,

Kalkofen

et al. 2023

IEEE Trans. Visual. Comput. Graphics

View full text Add to dashboard Cite

show abstract

“…Image-based rendering [3][1] methods are able to synthesize new views through the warping of images into 3D geometry, however rely on dense captures of the scene. Multiplane Images [11] allow view synthesis from a single view by applying depth estimation to slice the input image into multiple planes which are projected away from the camera. However, the movement of the camera is limited due to the need to inpaint [9] areas and the 2-dimensional nature of the planes revealing themselves.…”

Section: Related Workmentioning

confidence: 99%

“…However, they fall short of human rendering leaving room for research in achieving such edits to your everyday photo. Multiplane Imagery (MPI) [11] is another method that could be used for this purpose-they split the view into multiple parallel layers of depth to approximate the parallax. While decent for small viewpoint changes, these methods are not suitable for large camera motion, such as the one shown in Figure . 1.…”

Section: Introductionmentioning

confidence: 99%

Novel-View Synthesis of Human Tourist Photos

Freer

Jiang

et al. 2022

2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

View full text Add to dashboard Cite

Where a licence is displayed above, please note the terms and conditions of the licence govern your use of this document.When citing, please reference the published version. Take down policy While the University of Birmingham exercises care and attention in making items available there are rare occasions when an item has been uploaded in error or has been deemed to be commercially or otherwise sensitive.

show abstract

“…Once computed for a set of input images, novel views can be synthesized very efficiently by warping and blending the individual layers. One can attempt to predict MPIs from single input images [17,40]. Tucker and Snavely [40] train a network to estimate scale-invariant depth and require additional sparse point clouds to recover scale.…”

Section: Multiplane and Layered Depth Images (Mpis/ldis)mentioning

confidence: 99%