**Êtes-vous un étudiant de l'EPFL à la recherche d'un projet de semestre?**

Travaillez avec nous sur des projets en science des données et en visualisation, et déployez votre projet sous forme d'application sur GraphSearch.

Publication# GPU-accelerated Convex Multi-phase Image Segmentation

Xavier Bresson, Hadrien Copponnex, Arnaud Le Carvennec, Jean-Philippe Thiran, Subrahmanyam Venkata Ravi Mohana Sai Gorthi

2011

Rapport ou document de travail

2011

Rapport ou document de travail

Résumé

Image segmentation is a key area of research in computer vision. Recent advances facilitated reformulation of the non-convex multi-phase segmentation problem as a convex optimization problem (see for example [2, 4, 9, 10, 13, 16]). Recently, [3] proposed a new convex relaxation approach for a class of vector-valued minimization problems, and this approach is directly applicable to the widely used classical Mumford-Shah segmentation model [11]. While the approach in [3] provides the much deserved convexification, it achieves this at the expense of an increased computational complexity due to the increased dimensionality of the reformulated problem; however, the algorithm proposed in [3] can indeed profit from a parallelized implementation. In this paper, we present a GPU-based implementation of the convex formulation for Mumford-Shah piecewise constant multi-phase image segmentation algorithm proposed in [3]. The main goal of this paper is to provide insights into the way the algorithm has been parallelized in order to obtain good speedup. We present multi-phase segmentation results both on synthetic and real images. The speedup of GPU based implementation is evaluated on three different GPUs. For sufficiently large images, the speedup achieved on GTX 285 GPU is around 40, compared to an optimized CPU implementation. The speedups obtained from GPU-based implementation are quite satisfactory. We also made our CUDA code available online.

Official source

Cette page est générée automatiquement et peut contenir des informations qui ne sont pas correctes, complètes, à jour ou pertinentes par rapport à votre recherche. Il en va de même pour toutes les autres pages de ce site. Veillez à vérifier les informations auprès des sources officielles de l'EPFL.

Concepts associés

Chargement

Publications associées

Chargement

Publications associées (15)

Chargement

Chargement

Chargement

Concepts associés (10)

Vision par ordinateur

La vision par ordinateur est un domaine scientifique et une branche de l’intelligence artificielle qui traite de la façon dont les ordinateurs peuvent acquérir une compréhension de haut niveau à par

Segmentation d'image

La segmentation d'image est une opération de s consistant à détecter et rassembler les pixels suivant des critères, notamment d'intensité ou spatiaux, l'image apparaissant ainsi formée de régions uni

Dimension

Le terme dimension, du latin dimensio « action de mesurer », désigne d’abord chacune des grandeurs d’un objet : longueur, largeur et profondeur, épaisseur ou hauteur, ou encore son diamètre si c'est

Vasilis Kalofolias, Nathanaël Perraudin, Gilles Puy, Nauman Shahid, Pierre Vandergheynst

Mining useful clusters from high dimensional data has received significant attention of the computer vision and pattern recognition community in the recent years. Linear and non-linear dimensionality reduction has played an important role to overcome the curse of dimensionality. However, often such methods are accompanied with three different problems: high computational complexity (usually associated with the nuclear norm minimization), non-convexity (for matrix factorization methods) and susceptibility to gross corruptions in the data. In this paper we propose a principal component analysis (PCA) based solution that overcomes these three issues and approximates a low-rank recovery method for high dimensional datasets. We target the low-rank recovery by enforcing two types of graph smoothness assumptions, one on the data samples and the other on the features by designing a convex optimization problem. The resulting algorithm is fast, efficient and scalable for huge datasets with $\mathcal{O}(n \log(n))$ computational complexity in the number of data samples. It is also robust to gross corruptions in the dataset as well as to the model parameters. Clustering experiments on 7 benchmark datasets with different types of corruptions and background separation experiments on 3 video datasets show that our proposed model outperforms 10 state-of-the-art dimensionality reduction models.

Options are some of the most traded financial instruments and computing their price is a central task in financial mathematics and in practice. Consequently, the development of numerical algorithms for pricing options is an active field of research. In general, evaluating the price of a specific option relies on the properties of the stochastic model used for the underlying asset price. In this thesis we develop efficient and accurate numerical methods for option pricing in a specific class of models: polynomial models. They are a versatile tool for financial modeling and have useful properties that can be exploited for option pricing.
Significant challenges arise when developing option pricing techniques. For instance, the underlying model might have a high-dimensional parameter space. Furthermore, treating multi-asset options yields high-dimensional pricing problems. Therefore, the pricing method should be able to handle high dimensionality. Another important aspect is the efficiency of the algorithm: in real-world applications, option prices need to be delivered within short periods of time, making the algorithmic complexity a potential bottleneck. In this thesis, we address these challenges by developing option pricing techniques that are able to handle low and high-dimensional problems, and we propose complexity reduction techniques.
The thesis consists of four parts:
First, we present a methodology for European and American option pricing. The method uses the moments of the underlying price process to produce monotone sequences of lower and upper bounds of the option price. The bounds are obtained by solving a sequence of polynomial optimization problems. As the order of the moments increases, the bounds become sharper and eventually converge to the exact price under appropriate assumptions.
Second, we develop a fast algorithm for the incremental computation of nested block triangular matrix exponentials. This algorithm allows for an efficient incremental computation of the moment sequence of polynomial jump-diffusions. In other words, moments of order 0, 1, 2, 3... are computed sequentially until a dynamically evaluated criterion tells us to stop. The algorithm is based on the scaling and squaring technique and reduces the complexity of the pricing algorithms that require such an incremental moment computation.
Third, we develop a complexity reduction technique for high-dimensional option pricing. To this end, we first consider the option price as a function of model and payoff parameters. Then, the tensorized Chebyshev interpolation is used on the parameter space to increase the efficiency in computing option prices, while maintaining the required accuracy. The high dimensionality of the problem is treated by expressing the tensorized interpolation in the tensor train format and by deriving an efficient way, which is based on tensor completion, to approximate the interpolation coefficients.
Lastly, we propose a methodology for pricing single and multi-asset European options. The approach is a combination of Monte Carlo simulation and function approximation. We address the memory limitations that arise when treating very high-dimensional applications by combining the method with optimal sampling strategies and using a randomized algorithm to reduce the storage complexity of the approach.
The obtained numerical results show the effectiveness of the algorithms developed in this thesis.

Humans have the ability to learn. Having seen an object we can recognise it later. We can do this because our nervous system uses an efficient and robust visual processing and capabilities to learn from sensory input. On the other hand, designing algorithms to learn from visual data is a difficult task. More than fifty years ago, Rosenblatt proposed the perceptron algorithm. The perceptron learns from data examples a linear separation, which categorises the data in two classes. The algorithm served as a simple model of neuronal learning. Two further important ideas were added to the perceptron. First, to look for a maximal margin of separation. Second, to separate the data in a possibly high dimensional feature space, related nonlinearly to the initial space of the data, and allowing nonlinear separations. Important is that learning in the feature space can be performed implicitly and hence efficiently with the use of a kernel, a measure of similarity between two data points. The combination of these ideas led to the support vector machine, an efficient algorithm with high performance. In this thesis, we design an algorithm to learn the categorisation of data into multiple classes. This algorithm is applied to a real-time vision task, the recognition of human faces. Our algorithm can be seen as a generalisation of the support vector machine to multiple classes. It is shown how the algorithm can be efficiently implemented. To avoid a large number of small but time consuming updates of the variables limited accuracy computations are used. We prove a bound on the accuracy needed to find a solution. The proof motivates the use of a heuristic, which further increases efficiency. We derive a second implementation using a stochastic gradient descent method. This implementation is appealing as it has a direct interpretation and can be used in an online setting. Conceptually our approach differs from standard support vector approaches because examples can be rejected and are not necessarily attributed to one of the categories. This is natural in the context of a vision task. At any time, the sensory input can be something unseen before and hence cannot be recognised. Our visual data are images acquired with the recently developed adaptive vision sensor from CSEM. The vision sensor has two important features. First, like the human retina, it is locally adaptive to light intensity. Hence, the sensor has a high dynamic range. Second, the image gradient is computed on the sensor chip and is thus available directly from the sensor in real time. The sensor output is time encoded. The information about a strong local contrast is transmitted rst and the weakest contrast information at the end. To recognise faces, possibly moving in front of the camera, the sensor images have to be processed in a robust way. Representing images to exhibit local invariances is a common yet unsolved problem in computer vision. We develop the following representation of the sensor output. The image gradient information is decomposed into local histograms over contrast intensity. The histograms are local in position and direction of the gradient. Hence, the representation has local invariance properties to translation, rotation, and scaling. The histograms can be efficiently computed because the sensor output is already ordered with respect to the local contrast. Our support vector approach for multicategorical data uses the local histogram features to learn the recognition of faces. As recognition is time consuming, a face detection stage is used beforehand. We learn the detection features in an unsupervised manner using a specially designed optimisation procedure. The combined system to detect and recognise faces of a small group of individuals is efficient, robust, and reliable.