We present a novel approach to view synthesis using multiplane images (MPIs). Building on recent advances in learned gradient descent, our algorithm generates an MPI from a set of sparse camera viewpoints. The resulting method incorporates occlusion reasoning, improving performance on challenging scene features such as object boundaries, lighting reflections, thin structures, and scenes with high depth complexity. We show that our method achieves high-quality, state-of-the-art results on two datasets: the Kalantari light field dataset, and a new camera array dataset, Spaces, which we make publicly available.

We present a learning-based method to infer plausible high dynamic range (HDR), omnidirectional illumination given an unconstrained, low dynamic range (LDR) image from a mobile phone camera with a limited field of view (FOV). For training data, we collect videos of various reflective spheres placed within the camera's FOV, leaving most of the background unoccluded, leveraging that materials with diverse reflectance functions reveal different lighting cues in a single exposure. We train a deep neural network to regress from the LDR background image to HDR lighting by matching the LDR ground truth sphere images to those rendered with the predicted illumination using image-based relighting, which is differentiable. Our inference runs at interactive frame rates on a mobile device, enabling realistic rendering of virtual objects into real scenes for mobile mixed reality. Training on automatically exposed and white-balanced videos, we improve the realism of rendered objects compared to the state-of-the art methods for both indoor and outdoor scenes.

We present "The Relightables", a volumetric capture system for photorealistic and high quality relightable full-body performance capture. While significant progress has been made on volumetric capture systems, focusing on 3D geometric reconstruction with high resolution textures, much less work has been done to recover photometric properties needed for relighting. Results from such systems lack high-frequency details and the subject's shading is prebaked into the texture. In contrast, a large body of work has addressed relightable acquisition for image-based approaches, which photograph the subject under a set of basis lighting conditions and recombine the images to show the subject as they would appear in a target lighting environment. However, to date, these approaches have not been adapted for use in the context of a high-resolution volumetric capture system. Our method combines this ability to realistically relight humans for arbitrary environments, with the benefits of free-viewpoint volumetric capture and new levels of geometric accuracy for dynamic performances. Our subjects are recorded inside a custom geodesic sphere outfitted with 331 custom color LED lights, an array of high-resolution cameras, and a set of custom high-resolution depth sensors. Our system innovates in multiple areas: First, we designed a novel active depth sensor to capture 12.4 MP depth maps, which we describe in detail. Second, we show how to design a hybrid geometric and machine learning reconstruction pipeline to process the high resolution input and output a volumetric video. Third, we generate temporally consistent reflectance maps for dynamic performers by leveraging the information contained in two alternating color gradient illumination images acquired at 60Hz. Multiple experiments, comparisons, and applications show that The Relightables significantly improves upon the level of realism in placing volumetrically captured human performances into arbitrary CG scenes.

We present a novel technique to relight images of human faces by learning a model of facial reflectance from a database of 4D reflectance field data of several subjects in a variety of expressions and viewpoints. Using our learned model, a face can be relit in arbitrary illumination environments using only two original images recorded under spherical color gradient illumination. The output of our deep network indicates that the color gradient images contain the information needed to estimate the full 4D reflectance field, including specular reflections and high frequency details. While capturing spherical color gradient illumination still requires a special lighting setup, reduction to just two illumination conditions allows the technique to be applied to dynamic facial performance capture. We show side-by-side comparisons which demonstrate that the proposed system outperforms the state-of-the-art techniques in both realism and speed.

Lighting plays a central role in conveying the essence and depth of the subject in a portrait photograph. Professional photographers will carefully control the lighting in their studio to manipulate the appearance of their subject, while consumer photographers are usually constrained to the illumination of their environment. Though prior works have explored techniques for relighting an image, their utility is usually limited due to requirements of specialized hardware, multiple images of the subject under controlled or known illuminations, or accurate models of geometry and reflectance. To this end, we present a system for portrait relighting : a neural network that takes as input a single RGB image of a portrait taken with a standard cellphone camera in an unconstrained environment, and from that image produces a relit image of that subject as though it were illuminated according to any provided environment map. Our method is trained on a small database of 18 individuals captured under different directional light sources in a controlled light stage setup consisting of a densely sampled sphere of lights. Our proposed technique produces quantitatively superior results on our dataset's validation set compared to prior works, and produces convincing qualitative relighting results on a dataset of hundreds of real-world cellphone portraits. Because our technique can produce a 640 × 640 image in only 160 milliseconds, it may enable interactive user-facing photographic applications in the future.

The view synthesis problem---generating novel views of a scene from known imagery---has garnered recent attention due in part to compelling applications in virtual and augmented reality. In this paper, we explore an intriguing scenario for view synthesis: extrapolating views from imagery captured by narrow-baseline stereo cameras, including VR cameras and now-widespread dual-lens camera phones. We call this problem stereo magnification , and propose a learning framework that leverages a new layered representation that we call multiplane images (MPIs). Our method also uses a massive new data source for learning view extrapolation: online videos on YouTube. Using data mined from such videos, we train a deep network that predicts an MPI from an input stereo image pair. This inferred MPI can then be used to synthesize a range of novel views of the scene, including views that extrapolate significantly beyond the input baseline. We show that our method compares favorably with several recent view synthesis methods, and demonstrate applications in magnifying narrow-baseline stereo images.

We present a multi-view stereo reconstruction technique that directly produces a complete high-fidelity head model with consistent facial mesh topology. While existing techniques decouple shape estimation and facial tracking, our framework jointly optimizes for stereo constraints and consistent mesh parameterization. Our method is therefore free from drift and fully parallelizable for dynamic facial performance capture. We produce highly detailed facial geometries with artist-quality UV parameterization, including secondary elements such as eyeballs, mouth pockets, nostrils, and the back of the head. Our approach consists of deforming a common template model to match multi-view input images of the subject, while satisfying cross-view, cross-subject, and cross-pose consistencies using a combination of 2D landmark detection, optical flow, and surface and volumetric Laplacian regularization. Since the flow is never computed between frames, our method is trivially parallelized by processing each frame independently. Accurate rigid head pose is extracted using a PCA-based dimension reduction and denoising scheme. We demonstrate high-fidelity performance capture results with challenging head motion and complex facial expressions around eye and mouth regions. While the quality of our results is on par with the current state-of-the-art, our approach can be fully parallelized, does not suffer from drift, and produces face models with production-quality mesh topologies.

Creating and animating realistic 3D human faces is an important element of virtual reality, video games, and other areas that involve interactive 3D graphics. In this paper, we propose a system to generate photorealistic 3D blendshape-based face models automatically using only a single consumer RGB-D sensor. The capture and processing requires no artistic expertise to operate, takes 15 seconds to capture and generate a single facial expression, and approximately 1 minute of processing time per expression to transform it into a blendshape model. Our main contributions include a complete end-to-end pipeline for capturing and generating photorealistic blendshape models automatically and a registration method that solves dense correspondences between two face scans by utilizing facial landmarks detection and optical flows. We demonstrate the effectiveness of the proposed method by capturing different human subjects with a variety of sensors and puppeteering their 3D faces with real-time facial performance retargeting. The rapid nature of our method allows for just-in-time construction of a digital face. To that end, we also integrated our pipeline with a virtual reality facial performance capture system that allows dynamic embodiment of the generated faces despite partial occlusion of the user's real face by the head-mounted display.

Abstract We present a near‐instant method for acquiring facial geometry and reflectance using a set of commodity DSLR cameras and flashes. Our setup consists of twenty‐four cameras and six flashes which are fired in rapid succession with subsets of the cameras. Each camera records only a single photograph and the total capture time is less than the 67ms blink reflex. The cameras and flashes are specially arranged to produce an even distribution of specular highlights on the face. We employ this set of acquired images to estimate diffuse color, specular intensity, specular exponent, and surface orientation at each point on the face. We further refine the facial base geometry obtained from multi‐view stereo using estimated diffuse and specular photometric information. This allows final submillimeter surface mesostructure detail to be obtained via shape‐from‐specularity. The final system uses commodity components and produces models suitable for authoring high‐quality digital human characters.

We present a method for acquiring the per-pixel diffuse albedo, specular albedo, and surface normal maps of a subject at a single instant in time. The method is single-shot, requiring no optical flow, and per-pixel, making no assumptions regarding albedo statistics or surface connectivity. We photograph the subject inside a spherical illumination device emitting a static lighting pattern of vertically polarized RGB color gradients aligned with the XYZ axes, and horizontally polarized RGB color gradients inversely aligned with the XYZ axes. We capture simultaneous photographs using one of two possible setups: a single-view setup using a coaxially aligned camera pair with a polarizing beam splitter, and a multi-view stereo setup with different orientations of linear polarizing filters placed on the cameras, enabling high-quality geometry reconstruction. From this lighting we derive full-color diffuse albedo, single-channel specular albedo suitable for dielectric materials, and polarization-preserving surface normals which are free of corruption from subsurface scattering. We provide simple formulae to estimate the diffuse albedo, specular albedo, and surface normal maps in the single-view and multi-view cases and show error bounds which are smallfor many common subjects including faces.

Creating and animating a realistic 3D human face has been an important task in computer graphics. The capability of capturing the 3D face of a human subject and reanimate it quickly will find many applications in games, training simulations, and interactive 3D graphics. In this paper, we propose a system to capture photorealistic 3D faces and generate the blendshape models automatically using only a single commodity RGB-D sensor. Our method can rapidly generate a set of expressive facial poses from a single Microsoft Kinect and requires no artistic expertise on the part of the capture subject. The system takes only a matter of seconds to capture and produce a 3D facial pose and only requires 4 minutes of processing time to transform it into a blendshape model. Our main contributions include an end-to-end pipeline for capturing and generating face blendshape models automatically, and a registration method that solves dense correspondences between two face scans by utilizing facial landmark detection and optical flow. We demonstrate the effectiveness of the proposed method by capturing 3D facial models of different human subjects and puppeteering their models in an animation system with real-time facial performance retargeting.

We present a technique for synthesizing the effects of skin microstructure deformation by anisotropically convolving a high-resolution displacement map to match normal distribution changes in measured skin samples. We use a 10-micron resolution scanning technique to measure several in vivo skin samples as they are stretched and compressed in different directions, quantifying how stretching smooths the skin and compression makes it rougher. We tabulate the resulting surface normal distributions, and show that convolving a neutral skin microstructure displacement map with blurring and sharpening filters can mimic normal distribution changes and microstructure deformations. We implement the spatially-varying displacement map filtering on the GPU to interactively render the effects of dynamic microgeometry on animated faces obtained from high-resolution facial scans.

We present a process for rendering a realistic facial performance with control of viewpoint and illumination. The performance is based on one or more high-quality geometry and reflectance scans of an actor in static poses, driven by one or more video streams of a performance. We compute optical flow correspondences between neighboring video frames, and a sparse set of correspondences between static scans and video frames. The latter are made possible by leveraging the relightability of the static 3D scans to match the viewpoint(s) and appearance of the actor in videos taken in arbitrary environments. As optical flow tends to compute proper correspondence for some areas but not others, we also compute a smoothed, per-pixel confidence map for every computed flow, based on normalized cross-correlation. These flows and their confidences yield a set of weighted triangulation constraints among the static poses and the frames of a performance. Given a single artist-prepared face mesh for one static pose, we optimally combine the weighted triangulation constraints, along with a shape regularization term, into a consistent 3D geometry solution over the entire performance that is drift free by construction. In contrast to previous work, even partial correspondences contribute to drift minimization, for example, where a successful match is found in the eye region but not the mouth. Our shape regularization employs a differential shape term based on a spatially varying blend of the differential shapes of the static poses and neighboring dynamic poses, weighted by the associated flow confidences. These weights also permit dynamic reflectance maps to be produced for the performance by blending the static scan maps. Finally, as the geometry and maps are represented on a consistent artist-friendly mesh, we render the resulting high-quality animated face geometry and animated reflectance maps using standard rendering tools.

We present a novel technique for acquiring the geometry and spatially-varying reflectance properties of 3D objects by observing them under continuous spherical harmonic illumination conditions. The technique is general enough to characterize either entirely specular or entirely diffuse materials, or any varying combination across the surface of the object. We employ a novel computational illumination setup consisting of a rotating arc of controllable LEDs which sweep out programmable spheres of incident illumination during 1-second exposures. We illuminate the object with a succession of spherical harmonic illumination conditions, as well as photographed environmental lighting for validation. From the response of the object to the harmonics, we can separate diffuse and specular reflections, estimate world-space diffuse and specular normals, and compute anisotropic roughness parameters for each view of the object. We then use the maps of both diffuse and specular reflectance to form correspondences in a multiview stereo algorithm, which allows even highly specular surfaces to be corresponded across views. The algorithm yields a complete 3D model and a set of merged reflectance maps. We use this technique to digitize the shape and reflectance of a variety of objects difficult to acquire with other techniques and present validation renderings which match well to photographs in similar lighting.

ABSTRACT The linear blendshape technique has been intensively used for computer animation and games because of its simplicity and effectiveness. However, it cannot describe rotational deformations and deformations because of self collision or scene interaction. In this paper, we present a new technique to address these two major limitations by introducing physical‐based simulation to blendshapes. The proposed technique begins by constructing a mass–spring system for each blendshape target. Each system is initialized in its steady state by setting the rest length of each spring as the edge length of the corresponding target. To begin shape interpolation, we linearly interpolate the rest lengths of the springs according to a given interpolation factorα ∈ [0,1]. The interpolated shape is then generated by computing the equilibrium of the mass–spring system with the interpolated rest lengths. Results from our technique show physically plausible deformations even in the case of large rotations between blendshape targets. In addition, the new blendshape model is able to interact with other scene elements by introducing collision detection and handling to the mass–spring system. Copyright © 2012 John Wiley & Sons, Ltd.

We propose a novel method for single-shot photometric stereo by spectral multiplexing. The output of our method is a simultaneous per-pixel estimate of the surface normal and full-color reflectance. Our method is well suited to materials with varying color and texture, requires no time-varying illumination, and no high-speed cameras. Being a single-shot method, it may be applied to dynamic scenes without any need for optical flow. Our key contributions are a generalization of three-color photometric stereo to more than three color channels, and the design of a practical six-color-channel system using off-the-shelf parts.

Abstract We present a system for recording a live dynamic facial performance, capturing highly detailed geometry and spatially varying diffuse and specular reflectance information for each frame of the performance. The result is a reproduction of the performance that can be rendered from novel viewpoints and novel lighting conditions, achieving photorealistic integration into any virtual environment. Dynamic performances are captured directly, without the need for any template geometry or static geometry scans, and processing is completely automatic, requiring no human input or guidance. Our key contributions are a heuristic for estimating facial reflectance information from gradient illumination photographs, and a geometry optimization framework that maximizes a principled likelihood function combining multi‐view stereo correspondence and photometric stereo, using multi‐resolution belief propagation. The output of our system is a sequence of geometries and reflectance maps, suitable for rendering in off‐the‐shelf software. We show results from our system rendered under novel viewpoints and lighting conditions, and validate our results by demonstrating a close match to ground truth photographs.

We present a novel process for acquiring detailed facial geometry with high resolution diffuse and specular photometric information from multiple viewpoints using polarized spherical gradient illumination. Key to our method is a new pair of linearly polarized lighting patterns which enables multiview diffuse-specular separation under a given spherical illumination condition from just two photographs. The patterns -- one following lines of latitude and one following lines of longitude -- allow the use of fixed linear polarizers in front of the cameras, enabling more efficient acquisition of diffuse and specular albedo and normal maps from multiple viewpoints. In a second step, we employ these albedo and normal maps as input to a novel multi-resolution adaptive domain message passing stereo reconstruction algorithm to create high resolution facial geometry. To do this, we formulate the stereo reconstruction from multiple cameras in a commonly parameterized domain for multiview reconstruction. We show competitive results consisting of high-resolution facial geometry with relightable reflectance maps using five DSLR cameras. Our technique scales well for multiview acquisition without requiring specialized camera systems for sensing multiple polarization states.

Spectral multiplexing allows multiple channels of information to be captured simultaneously, using readily available color cameras. Information may be multiplexed across the color channels of a camera by use of colored lights (e.g. [Woodham 1980; Hernandez and Vogiatzis 2010]) or colored filters (e.g. [Bando et al. 2008]). We propose a novel method for single-shot photometric stereo by spectral multiplexing. The output of our method is a simultaneous per-pixel estimate of the surface normal and full-color reflectance. Our method is well suited to materials with varying color and texture, requires no time-varying illumination, and no high-speed cameras. Being a single-shot method, it may be applied to dynamic scenes without any need for optical flow. Our key contributions are a generalization of three-color photometric stereo to multiple (more than three) color channels, and the design of a practical six-color-channel system using off-the-shelf parts only.

We present a set of algorithms and an associated display system capable of producing correctly rendered eye contact between a three-dimensionally transmitted remote participant and a group of observers in a 3D teleconferencing system. The participant's face is scanned in 3D at 30Hz and transmitted in real time to an autostereoscopic horizontal-parallax 3D display, displaying him or her over more than a 180° field of view observable to multiple observers. To render the geometry with correct perspective, we create a fast vertex shader based on a 6D lookup table for projecting 3D scene vertices to a range of subject angles, heights, and distances. We generalize the projection mathematics to arbitrarily shaped display surfaces, which allows us to employ a curved concave display surface to focus the high speed imagery to individual observers. To achieve two-way eye contact, we capture 2D video from a cross-polarized camera reflected to the position of the virtual participant's eyes, and display this 2D video feed on a large screen in front of the real participant, replicating the viewpoint of their virtual self. To achieve correct vertical perspective, we further leverage this image to track the position of each audience member's eyes, allowing the 3D display to render correct vertical perspective for each of the viewers around the device. The result is a one-to-many 3D teleconferencing system able to reproduce the effects of gaze, attention, and eye contact generally missing in traditional teleconferencing systems.