State-of-the-art motion estimation algorithms suffer from three major problems: Poorly textured regions, occlusions and small scale image structures. Based on the Gestalt principles of grouping we propose to incorpora...
详细信息
ISBN:
(纸本)9781424469840
State-of-the-art motion estimation algorithms suffer from three major problems: Poorly textured regions, occlusions and small scale image structures. Based on the Gestalt principles of grouping we propose to incorporate a low level image segmentation process in order to tackle these problems. Our new motion estimation algorithm is based on non-local total variation regularization which allows us to integrate the low level image segmentation process in a unified variational framework. Numerical results on the Middlebury optical flow benchmark data set demonstrate that we can cope with the aforementioned problems.
In this paper, we propose a computationally efficient variant of elitist covariance matrix evolution strategy for continuous local search in high dimensional space. It focuses on searching in a low-dimensional subspac...
详细信息
ISBN:
(纸本)9781728121536
In this paper, we propose a computationally efficient variant of elitist covariance matrix evolution strategy for continuous local search in high dimensional space. It focuses on searching in a low-dimensional subspace expanded by a small number of promising search directions. This leads to the linear internal computational complexity of each iteration, which enables the algorithm to scale to high dimensional problems. We conduct comprehensive experiments to evaluate the parameter sensitivity and the algorithm's performance. The experimental results validate that the proposed algorithm reduces the running time by a factor of ten, and it can be easily scaled up to n > 1000 on a set of commonly used test functions.
The concept of scene graphs is widely used in computergraphics to structure graphics-related entities, e.g. geometry, visual attributes as well as abstract data related to certain application requirements like object...
详细信息
ISBN:
(纸本)9783980487481
The concept of scene graphs is widely used in computergraphics to structure graphics-related entities, e.g. geometry, visual attributes as well as abstract data related to certain application requirements like object identifiers or manufacturing details. This paper presents a new method to incorporate General Purpose graphics Programming Unit (GPGPU)-functionality into scene graph APIs. We define specific scene graph nodes in order to realize a flexible integration of GPU functionality at various levels of granularity without violating the programming paradigm inherent to scene graphs. We focus on current and upcoming compute APIs like CUDA, which are designed for GPGPU purposes. We further present the osgCompute framework that implements our concept and is based on the OpenSceneGraph API. CUDA is integrated into osgCompute via osgCuda. Our method is flexible in the sense that other compute APIs could be used instead. The advantages of our concept and of osgCuda are demonstrated by presenting examples with different processing requirements.
In many systems hierarchical data structures are used to accelerate the access to spatial data, whereas other applications use a hierarchical volumetric data structure to implicitly represent 3D objects. We introduce ...
详细信息
ISBN:
(纸本)9783980487481
In many systems hierarchical data structures are used to accelerate the access to spatial data, whereas other applications use a hierarchical volumetric data structure to implicitly represent 3D objects. We introduce Dynamic Volume Trees (DVT), i.e. an adaptive hierarchical volume data structure which can be modified in real-time. The online capability is achieved by realizing both the hierarchical data structure and the manipulation of the structure solely on the graphics Processing Unit (GPU). Even though we focus on the representation of highly-detailed volumetric scenes which may be gathered by sensors, e.g. in the fields of robotics and remote sensing applications, DVTs may easily be used to reference different data such as object references. The data is organized in a hierarchical kd-tree-like structure which provides a compact storage of multi-resolution volumes with no redundant memory consumption. Boolean operations are supported, i.e. sub-volumes can be efficiently merged (set union) and removed (set subtraction) with nearly arbitrary resolution. Additionally, a tree optimization is realized in order to improve the performance online. Furthermore, we present two approaches to render the data structure. The power and robustness of DVTs are demonstrated by a multi-resolution volume drawing example.
Background: It is unclear, whether photoplethysmography (PPG) waveforms from wearable devices can differentiate between supraventricular and ventricular arrhythmias. We assessed, whether a neural network-based classif...
详细信息
The complexity of a building façade provides challenge for window detection algorithms. Especially in the gradient projection approach, the presence of gradients outside window areas significantly reduces the qua...
详细信息
In this paper, we present a method for window detection, robust enough to process complex façades of historical buildings. This method is able to provide results even for facades under severe perspective distorti...
详细信息
This paper presents a novel system for interactive visualization and manipulation of medical datasets for surgery planning based on a hybrid VR / Tablet PC user interface. The goal of the system is to facilitate effic...
详细信息
Robust integration of range images is an important task for building high-quality 3D models. Since range images, and in particular range maps from stereo vision, may have a substantial amount of outliers, any integrat...
详细信息
ISBN:
(纸本)9781424416301
Robust integration of range images is an important task for building high-quality 3D models. Since range images, and in particular range maps from stereo vision, may have a substantial amount of outliers, any integration approach aiming at high-quality models needs an increased level of robustness. Additionally, a certain level of regularization is required to obtain smooth surfaces. Computational efficiency and global convergence are further preferable properties. The contribution of this paper is a unified framework to solve all these issues. Our method is based on minimizing an energy functional consisting of a total variation (TV) regularization force and an L-1 data fidelity term. We present a novel and efficient numerical scheme, which combines the duality principle for the TV term with a point-wise optimization step. We demonstrate the superior performance of our algorithm on the well-known Middlebury multi-view database and additionally on real-world multi-view images.
暂无评论