Graph Matching Techniques for Computer Vision

Many computer vision applications require a comparison between two objects, or between an object and a reference model. When the objects or the scenes are represented by graphs, this comparison can be performed using some form of graph matching. The aim of this chapter is to introduce the main graph matching techniques that have been used for computer vision, and to relate each application with the techniques that are most suited to it.

Graph Embedding Using Dissimilarities with Applications in Classification

The domain of graphs contains only little mathematical structure. That is, most of the basic mathematical operations, actually required by many standard computer vision and pattern recognition algorithms, are not available for graphs. One of the few mathematical concepts that has been successfully transferred from the vector space to the graph domain is distance computation between graphs, commonly referred to as graph matching. Yet, distance-based pattern recognition is basically limited to nearest-neighbor classification. The present chapter reviews a novel approach for graph embedding in vector spaces built upon the concept of graph matching. The key-idea of the proposed embedding method is to use the distances of an input graph to a number of training graphs, termed prototypes, as vectorial description of the graph. That is, all graph matching procedures proposed in the literature during the last decades can be employed in this embedding framework. The rationale for such a graph embedding is to bridge the gap between the high representational power and flexibility of graphs and the large amount of algorithms available for object representations in terms of feature vectors. Hence, the proposed framework can be considered a contribution towards unifying the domains of structural and statistical pattern recognition.

Certain and Uncertain Triangulation in Multiple Camera Vision Systems via LMIs

Triangulation is a fundamental problem in computer vision that consists of estimating the 3D position of a point of the scene from the estimates of its image projections on some cameras and from the estimates of the projection matrices of these cameras. This chapter addresses multiple view L2 triangulation, i.e. triangulation for vision systems with a generic number of cameras where the sought 3D point is searched by minimizing the L2 norm of the image reprojection error. The authors consider the standard case where estimates of all the image points are available (referring to such a case as certain triangulation), and consider also the case where some of such estimates are not available for example due to occlusions (referring to such a case as uncertain triangulation). In the latter case, it is supposed that the unknown image points belong to known regions such as line segments or ellipses. For these problems, the authors propose a unified methodology that exploits the fundamental matrices among the views and provides a candidate 3D point through the solution of a convex optimization problem based on linear matrix inequalities (LMIs). Moreover, the chapter provides a simple condition that allows one to immediately establish whether the found 3D point is optimal. Various examples with synthetic and real data illustrate the proposed technique, showing in particular that the obtained 3D point is almost always optimal in practice, and that its computational time is indeed small.

The Kolmogorov Spline Network for Image Processing

In 1900, Hilbert stated that high order equations cannot be solved by sums and compositions of bivariate functions. In 1957, Kolmogorov proved this hypothesis wrong and presented his superposition theorem (KST) that allowed for writing every multivariate functions as sums and compositions of univariate functions. Sprecher has proposed in (Sprecher, 1996) and (Sprecher, 1997) an algorithm for exact univariate function reconstruction. Sprecher explicitly describes construction methods for univariate functions and introduces fundamental notions for the theorem comprehension (such as tilage). Köppen has presented applications of this algorithm to image processing in (Köppen, 2002) and (Köppen & Yoshida, 2005). The lack of flexibility of this scheme has been pointed out and another solution which approximates the univariate functions has been considered. More specifically, it has led us to consider Igelnik and Parikh’s approach, known as the KSN which offers several perspectives of modification of the univariate functions as well as their construction. This chapter will focus on the presentation of Igelnik and Parikh’s Kolmogorov Spline Network (KSN) for image processing and detail two applications: image compression and progressive transmission. Precisely, the developments presented in this chapter include: (1)Compression: the authors study the reconstruction quality using univariate functions containing only a fraction of the original image pixels. To improve the reconstruction quality, they apply this decomposition on images of details obtained by wavelet decomposition. The authors combine this approach into the JPEG 2000 encoder, and show that the obtained results improve JPEG 2000 compression scheme, even at low bitrates. (2)Progressive Transmission: the authors propose to modify the generation of the KSN. The image is decomposed into univariate functions that can be transmitted one after the other to add new data to the previously transmitted functions, which allows to progressively and exactly reconstruct the original image. They evaluate the transmission robustness and provide the results of the simulation of a transmission over packet-loss channels.

Replicating the Role of the Human Retina for a Cortical Visual Neuroprosthesis

Neuroengineering is an emerging research field combining the latest findings from neuroscience with developments in a variety of engineering disciplines to create artificial devices, mainly for therapeutical purposes. In this chapter, an application of this field to the development of a visual neuroprosthesis for the blind is described. Electrical stimulation of the visual cortex in blind subjects elicits the perception of visual sensations called phosphenes, a finding that encourages the development of future electronic visual prostheses. However, direct stimulation of the visual cortex would miss a significant degree of image processing that is carried out by the retina. The authors describe a biologically-inspired retina-like processor designed to drive the implanted stimulator using visual inputs from one or two cameras. This includes dynamic response modeling with minimal latency. The outputs of the retina-like processor are comparable to those recorded in biological retinas that are exposed to the same stimuli and allow estimation of the original scene.

Face Recognition with Active Appearance Model (AAM)

Recognition of human faces has been a fascinating subject in research field for many years. It is considered a multidisciplinary field because it includes understanding different domains such as psychology, neuroscience, computer vision, artificial intelligence, mathematics, and many others. Human face perception is intriguing and draws our attention because we accomplish the task so well that we hope to one day witness a machine performing the same task in a similar or better way. This chapter aims to provide a systematic and practical approach regarding to one of the most current techniques applied on face recognition, known as AAM (Active Appearance Model). AAM method is addressed considering 2D face processing only. This chapter doesn’t cover the entire theme, but offers to the reader the necessary tools to construct a consistent and productive pathway toward this involving subject.

Facial Image Processing in Computer Vision

The application of computer vision in face processing remains an important research field. The aim of this chapter is to provide an up-to-date review of research efforts of computer vision scientist in facial image processing, especially in the areas of entertainment industry, surveillance, and other human computer interaction applications. To be more specific, this chapter reviews and demonstrates the techniques of visible facial analysis, regardless of specific application areas. First, the chapter makes a thorough survey and comparison of face detection techniques. It provides some demonstrations on the effect of computer vision algorithms and colour segmentation on face images. Then, it reviews the facial expression recognition from the psychological aspect (Facial Action Coding System, FACS) and from the computer animation aspect (MPEG-4 Standard). The chapter also discusses two popular existing facial feature detection techniques: Gabor feature based boosted classifiers and Active Appearance Models, and demonstrate the performance on our in-house dataset. Finally, the chapter concludes with the future challenges and future research direction of facial image processing.

Security Applications Using Computer Vision

Computer vision plays a significant role in a wide range of homeland security applications. The homeland security applications include: port security (cargo inspection), facility security (embassy, power plant, bank), and surveillance (military or civilian), et cetera. Video surveillance cameras are placed in offices, hospitals, banks, ports, parking lots, parks, stadiums, malls, train stations, airports, et cetera. The challenge is not for acquiring surveillance data from these video cameras, but for identifying what is valuable, what can be ignored, and what demands immediate attention. Computer vision systems attempt to construct meaningful and explicit descriptions of the environment or scene captured in an image. A few Computer Vision based security applications are presented here for securing building facility, railroad (Objects on railroad, and red signal detection), and roads.

Feature Extraction Methods for Intrusion Detection Systems

Intrusion Detection Systems (IDSs) have become an important security tool for managing risk and an indispensable part of overall security architecture. An IDS is considered as a pattern recognition system, in which feature extraction is an important pre-processing step. The feature extraction process consists of feature construction and feature selection . The quality of the feature construction and feature selection algorithms is one of the most important factors that affects the effectiveness of an IDS. Achieving reduction of the number of relevant traffic features without negative effect on classification accuracy is a goal that largely improves the overall effectiveness of the IDS. Most of the feature construction as well as feature selection works in intrusion detection practice is still carried through manually by utilizing domain knowledge. For automatic feature construction and feature selection, the filter, wrapper, and embedded methods from machine learning are frequently applied. This chapter provides an overview of various existing feature construction and feature selection methods for intrusion detection systems. A comparison between those feature selection methods is performed in the experimental part.

Building a Multiple Object Tracking System with Occlusion Handling in Surveillance Videos

Video tracking systems are increasingly used day in and day out in various applications such as surveillance, security, monitoring, and robotic vision. In this chapter, the authors propose a novel multiple objects tracking system in video sequences that deals with occlusion issues. The proposed system is composed of two components: An improved KLT tracker, and a Kalman filter. The improved KLT tracker uses the basic KLT tracker and an appearance model to track objects from one frame to another and deal with partial occlusion. In partial occlusion, the appearance model (e.g., a RGB color histogram) is used to determine an object’s KLT features, and the authors use these features for accurate and robust tracking. In full occlusion, a Kalman filter is used to predict the object’s new location and connect the trajectory parts. The system is evaluated on different videos and compared with a common tracking system.