MoReLab: A Software for User-Assisted 3D Reconstruction.

We present MoReLab, a tool for user-assisted 3D reconstruction. This reconstruction requires an understanding of the shapes of the desired objects. Our experiments demonstrate that existing Structure from Motion (SfM) software packages fail to estimate accurate 3D models in low-quality videos due to several issues such as low resolution, featureless surfaces, low lighting, etc. In such scenarios, which are common for industrial utility companies, user assistance becomes necessary to create reliable 3D models. In our system, the user first needs to add features and correspondences manually on multiple video frames. Then, classic camera calibration and bundle adjustment are applied. At this point, MoReLab provides several primitive shape tools such as rectangles, cylinders, curved cylinders, etc., to model different parts of the scene and export 3D meshes. These shapes are essential for modeling industrial equipment whose videos are typically captured by utility companies with old video cameras (low resolution, compression artifacts, etc.) and in disadvantageous lighting conditions (low lighting, torchlight attached to the video camera, etc.). We evaluate our tool on real industrial case scenarios and compare it against existing approaches. Visual comparisons and quantitative results show that MoReLab achieves superior results with regard to other user-interactive 3D modeling tools.

Characterization of Nanoparticles in Diverse Mixtures Using Localized Surface Plasmon Resonance and Nanoparticle Tracking by Dark-Field Microscopy with Redox Magnetohydrodynamics Microfluidics.

Redox magnetohydrodynamics (RMHD) microfluidics is coupled with dark-field microscopy (DFM) to offer high-throughput single-nanoparticle (NP) differentiation in situ and operando in a flowing mixture by localized surface plasmon resonance (LSPR) and tracking of NPs. The color of the scattered light allows visualization of the NPs below the diffraction limit. Their Brownian motion in 1-D superimposed on and perpendicular to the RMHD trajectory yields their diffusion coefficients. LSPR and diffusion coefficients provide two orthogonal modalities for characterization where each depends on a particle's material composition, shape, size, and interactions with the surrounding medium. RMHD coupled with DFM was demonstrated on a mixture of 82 ± 9 nm silver and 140 ± 10 nm gold-coated silica nanospheres. The two populations of NPs in the mixture were identified by blue/green and orange/red LSPR and their scattering intensity, respectively, and their sizes were further evaluated based on their diffusion coefficients. RMHD microfluidics facilitates high-throughput analysis by moving the sample solution across the wide field of view absent of physical vibrations within the experimental cell. The well-controlled pumping allows for a continuous, reversible, and uniform flow for precise and simultaneous NP tracking of the Brownian motion. Additionally, the amounts of nanomaterials required for the analysis are minimized due to the elimination of an inlet and outlet. Several hundred individual NPs were differentiated from each other in the mixture flowing in forward and reverse directions. The ability to immediately reverse the flow direction also facilitates re-analysis of the NPs, enabling more precise sizing.

Unexpectedly High Capacitance of the Metal Nanoparticle/Water Interface: Molecular-Level Insights into the Electrical Double Layer.

The electrical double-layer plays a key role in important interfacial electrochemical processes from catalysis to energy storage and corrosion. Therefore, understanding its structure is crucial for the progress of sustainable technologies. We extract new physico-chemical information on the capacitance and structure of the electrical double-layer of platinum and gold nanoparticles at the molecular level, employing single nanoparticle electrochemistry. The charge storage ability of the solid/liquid interface is larger by one order-of-magnitude than predicted by the traditional mean-field models of the double-layer such as the Gouy-Chapman-Stern model. Performing molecular dynamics simulations, we investigate the possible relationship between the measured high capacitance and adsorption strength of the water adlayer formed at the metal surface. These insights may launch the active tuning of solid-solvent and solvent-solvent interactions as an innovative design strategy to transform energy technologies towards superior performance and sustainability.

Turning a Smartphone Selfie Into a Studio Portrait.

We introduce a novel algorithm that turns a flash selfie taken with a smartphone into a studio-like photograph with uniform lighting. Our method uses a convolutional neural network trained on a set of pairs of photographs acquired in a controlled environment. For each pair, we have one photograph of a subject's face taken with the camera flash enabled and another one of the same subject in the same pose illuminated using a photographic studio-lighting setup. We show how our method can amend lighting artifacts introduced by a close-up camera flash, such as specular highlights, shadows, and skin shine.

PhotoCloud: Interactive remote exploration of joint 2D and 3D datasets.

PhotoCloud is a real-time client-server system for interactive visualization and exploration of large datasets comprising thousands of calibrated 2D photographs of a scene and a complex 3D description of the scene. The system isn't tailored to any specific data acquisition process; it aims at generality and flexibility. PhotoCloud achieves scalability through a multiresolution dynamic hierarchical representation of the data, which is remotely stored and accessed by the client through an efficient cache system. The system includes a compact image browser and a multiresolution model renderer. PhotoCloud employs iconic visualization of the images in the 3D space and projects images onto the 3D scene on the fly. Users can navigate the 2D and 3D spaces with smooth, integrated, seamless transitions between them. A study with differently skilled users confirms PhotoCloud's effectiveness and communication power. The Web extras at http://www.youtube.com/playlist?list=PLHJB2bhmgB7cmYD0ST9CEDMRv1JlX4xPH are videos demonstrating PhotoCloud, a real-time client-server system for interactive exploration of large datasets comprising 2D photos and 3D models.

Efficient and flexible sampling with blue noise properties of triangular meshes.

This paper deals with the problem of taking random samples over the surface of a 3D mesh describing and evaluating efficient algorithms for generating different distributions. We discuss first the problem of generating a Monte Carlo distribution in an efficient and practical way avoiding common pitfalls. Then, we propose Constrained Poisson-disk sampling, a new Poisson-disk sampling scheme for polygonal meshes which can be easily tweaked in order to generate customized set of points such as importance sampling or distributions with generic geometric constraints. In particular, two algorithms based on this approach are presented. An in-depth analysis of the frequency characterization and performance of the proposed algorithms are also presented and discussed.

Flow-based local optimization for image-to-geometry projection.

The projection of a photographic data set on a 3D model is a robust and widely applicable way to acquire appearance information of an object. The first step of this procedure is the alignment of the images on the 3D model. While any reconstruction pipeline aims at avoiding misregistration by improving camera calibrations and geometry, in practice a perfect alignment cannot always be reached. Depending on the way multiple camera images are fused on the object surface, remaining misregistrations show up either as ghosting or as discontinuities at transitions from one camera view to another. In this paper we propose a method, based on the computation of Optical Flow between overlapping images, to correct the local misalignment by determining the necessary displacement. The goal is to correct the symptoms of misregistration, instead of searching for a globally consistent mapping, which might not exist. The method scales up well with the size of the data set (both photographic and geometric) and is quite independent of the characteristics of the 3D model (topology cleanliness, parametrization, density). The method is robust and can handle real world cases that have different characteristics: low level geometric details and images that lack enough features for global optimization or manual methods. It can be applied to different mapping strategies, such as texture or per-vertex attribute encoding.

An interactive local flattening operator to support digital investigations on artwork surfaces.

Analyzing either high-frequency shape detail or any other 2D fields (scalar or vector) embedded over a 3D geometry is a complex task, since detaching the detail from the overall shape can be tricky. An alternative approach is to move to the 2D space, resolving shape reasoning to easier image processing techniques. In this paper we propose a novel framework for the analysis of 2D information distributed over 3D geometry, based on a locally smooth parametrization technique that allows us to treat local 3D data in terms of image content. The proposed approach has been implemented as a sketch-based system that allows to design with a few gestures a set of (possibly overlapping) parameterizations of rectangular portions of the surface. We demonstrate that, due to the locality of the parametrization, the distortion is under an acceptable threshold, while discontinuities can be avoided since the parametrized geometry is always homeomorphic to a disk. We show the effectiveness of the proposed technique to solve specific Cultural Heritage (CH) tasks: the analysis of chisel marks over the surface of a unfinished sculpture and the local comparison of multiple photographs mapped over the surface of an artwork. For this very difficult task, we believe that our framework and the corresponding tool are the first steps toward a computer-based shape reasoning system, able to support CH scholars with a medium they are more used to.

Automatic construction of quad-based subdivision surfaces using Fitmaps.

We present an automatic method to produce a Catmull-Clark subdivision surface that fits a given input mesh. Its control mesh is coarse and adaptive, and it is obtained by simplifying an initial mesh at high resolution. Simplification occurs progressively via local operators and addresses both quality of surface and faithfulness to the input shape throughout the whole process. The method is robust and performs well on rather complex shapes. Displacement mapping or normal mapping can be applied to approximate the input shape arbitrarily well.

Browsing large image datasets through Voronoi diagrams.

Conventional browsing of image collections use mechanisms such as thumbnails arranged on a regular grid or on a line, often mounted over a scrollable panel. However, this approach does not scale well with the size of the datasets (number of images). In this paper, we propose a new thumbnail-based interface to browse large collections of images. Our approach is based on weighted centroidal anisotropic Voronoi diagrams. A dynamically changing subset of images is represented by thumbnails and shown on the screen. Thumbnails are shaped like general polygons, to better cover screen space, while still reflecting the original aspect ratios or orientation of the represented images. During the browsing process, thumbnails are dynamically rearranged, reshaped and rescaled. The objective is to devote more screen space (more numerous and larger thumbnails) to the parts of the dataset closer to the current region of interest, and progressively lesser away from it, while still making the dataset visible as a whole. During the entire process, temporal coherence is always maintained. GPU implementation easily guarantees the frame rates needed for fully smooth interactivity.

Solid-texture synthesis: a survey.

Solid textures are an efficient way to compactly represent 3D objects' external and internal appearance, providing practical advantages over classic 2D texturing. Two main methods exist for synthesizing solid textures. Procedural methods obtain colors through functions that algorithmically encode the texture's appearance and structural properties. Example-based methods capture and replicate the appearance as described by a set of input exemplars. These methods can also be classified as boundary independent or boundary dependent. For boundary-independent methods, the shape of the object to be textured is irrelevant, and texture information can be freely generated for each point in the space. Boundary-dependent methods conform the synthesis process to the object's actual shape so that they can exploit this information to orient and guide texture generation. This article reviews the different methodologies' strengths and weaknesses, the classes of appearances they can successfully synthesize, and failure cases. In particular, it focuses on boundary-independent methods' advantages and drawbacks compared to boundary-dependent methods.

Almost isometric mesh parameterization through abstract domains.

In this paper, we propose a robust, automatic technique to build a global hi-quality parameterization of a two-manifold triangular mesh. An adaptively chosen 2D domain of the parameterization is built as part of the process. The produced parameterization exhibits very low isometric distortion, because it is globally optimized to preserve both areas and angles. The domain is a collection of equilateral triangular 2D regions enriched with explicit adjacency relationships (it is abstract in the sense that no 3D embedding is necessary). It is tailored to minimize isometric distortion, resulting in excellent parameterization qualities, even when meshes with complex shape and topology are mapped into domains composed of a small number of large continuous regions. Moreover, this domain is, in turn, remapped into a collection of 2D square regions, unlocking many advantages found in quad-based domains (e.g., ease of packing). The technique is tested on a variety of cases, including challenging ones, and compares very favorably with known approaches. An open-source implementation is made available.

Ambient occlusion and edge cueing to enhance real time molecular visualization.

The paper presents a set of combined techniques to enhance the real-time visualization of simple or complex molecules (up to order of 106 atoms) space fill mode. The proposed approach includes an innovative technique for efficient computation and storage of ambient occlusion terms, a small set of GPU accelerated procedural impostors for space-fill and ball-and-stick rendering, and novel edge-cueing techniques. As a result, the user's understanding of the three-dimensional structure under inspection is strongly increased (even for still images), while the rendering still occurs in real time.

Selective refinement queries for volume visualization of unstructured tetrahedral meshes.

In this paper, we address the problem of the efficient visualization of large irregular volume data sets by exploiting a multiresolution model based on tetrahedral meshes. Multiresolution models, also called Level-Of-Detail (LOD) models, allow encoding the whole data set at a virtually continuous range of different resolutions. We have identified a set of queries for extracting meshes at variable resolution from a multiresolution model, based on field values, domain location, or opacity of the transfer function. Such queries allow trading off between resolution and speed in visualization. We define a new compact data structure for encoding a multiresolution tetrahedral mesh built through edge collapses to support selective refinement efficiently and show that such a structure has a storage cost from 3 to 5.5 times lower than standard data structures used for tetrahedral meshes. The data structures and variable resolution queries have been implemented together with state-of-the art visualization techniques in a system for the interactive visualization of three-dimensional scalar fields defined on tetrahedral meshes. Experimental results show that selective refinement queries can support interactive visualization of large data sets.