Mihaela Scuturici

Real Time Foreground-Background Segmentation Using a Modified Codebook Model

Real time segmentation of scene into objects and background is really important and represents an initial step of object tracking. Starting from the codebook method [4] we propose some modifications which show significant improvements in most of the normal and also difficult conditions. We include parameter of frequency for accessing, deleting, matching and adding codewords in codebook or to move cache codewords into codebook. We also propose an evaluation method in order to objectively compare several segmentation techniques, based on receiver operating characteristic (ROC) analysis and on precision and recall method. We propose to summarize the quality factor of a method by a single value based on a weighted Euclidean distance or on a harmonic mean between two related characteristics.

Topological representation model for image database query

Search algorithms in image databases usually return k nearest neighbours (kNN) of an image according to a similarity measure. This approach presents some anomalies and is based on assumptions that are not always satisfied. We have examined the causes of these anomalies and we have concluded that image query models have to exploit topological properties rather than the similarity degree. This paper proposes a topological model based on neighbourhood graphs built on automatically extracted image features. Each image is represented as a feature vector in R p and stands for a node in the neighbourhood graph. The graph exploration corresponds to database browsing, the neighbours of a node represent similar images. In order to perform query by example, the query image is represented as a R p feature vector and inserted in the graph by locally updating the neighbourhood graph. The topology of an image database is more informative than a similarity measure usually applied in content based image retrieval, as proved by our experiments. A prototype of a visualization and query tool called Smart Image Query (SIQ) is also introduced.

Inter-camera color calibration for object re-identification and tracking

Object re-identification and tracking in non-overlapping cameras is a challenging problem due to the variation of the objects appearance, linked to the different view angle, distance and color variation in different cameras. We present a computationally efficient real time human tracking algorithm, which can track objects inside the field of view (FOV) of a camera, re-identify objects that exit and then return in a same or in a different camera FOV. Object appearance in several cameras may be very different due to illumination conditions, camera gain, focus, focal length etc. Therefore inter-camera color calibration is important for object re-identification before applying object recognition features. We compare different existing color calibration methods and evaluate their color brightness transfer function (BTF) using Receiver Operating Characteristic (ROC) curve. We propose some modifications in cumulative brightness transfer function (CBTF) which significantly improve the objects re-identification in non-overlapping multi camera environment.

A combined motion and appearance model for human tracking in multiple cameras environment

The aim of this paper is to present an algorithm for multiple object tracking and video summarization in a scene filmed by one or several cameras. We propose a computationally efficient real time human tracking algorithm, which can 1) track objects inside the field of view (FOV) of a camera even in case of occlusions; 2) recognize objects that quit and then return on a cameras FOV; 3) recognize objects passing through different cameras FOV. We propose a simple 1-D appearance model, called vertical feature (VF), view and size invariant, which is stored in a database in order to help object recognition. We combine it with other motion features like position and velocity for real-time tracking. We find the k closest matches of current object and select the one whose predicted position is closest to the current object position. Our algorithm shows good capabilities for objects tracking even with the change of object view angle and also with the partial change of shape. We compare our algorithm with appearance based and motion based algorithms and show the advantage of a combined approach.

A New Pixel-Based Quality Measure for Segmentation Algorithms Integrating Precision, Recall and Specificity

There are several approaches for performance evaluation of image processing algorithms in video-based surveillance systems: Precision/ Recall, Receiver Operator Characteristics ROC, F-measure, Jaccard Coefficient, etc. These measures can be used to find good values for input parameters of image segmentation algorithms. Different measures can give different values of these parameters, considered as optimal by one criterion, but not by another. Most of the times, the measures are expressed as a compromise between two of the three aspects that are important for a quality assessment: Precision, Recall and Specificity. In this paper, we propose a new 3-dimensional measure D prs , which takes into account all of the three aspects. It can be considered as a 3D generalization of 2D ROC analysis and Precision/Recall curves. To estimate the impact of parameters on the quality of the segmentation, we study the behavior of this measure and compare it with several classical measures. Both objective and subjective evaluations confirm that our new measure allows to determine more stable parameters than classical criteria, and to obtain better segmentations of images.

Tracking 3D free form object in video sequence

In this paper we describe an original method for the 3D free form object tracking in monocular vision. The main contribution of this article is the use of the skeleton of an object in order to recognize, locate and track this object in real time. Indeed, the use of this kind of representation made it possible to avoid difficulties related to the absence of prominent elements in free form objects (which makes the matching process easier). The skeleton is a lower dimension representation of the object, it is homotopic and it has a graph structure. This allowed us to use powerful tools of the graph theory in order to perform matching between scene objects and models (recognition step). Thereafter, we used skeleton extremities as interest points for the tracking. Keywords: Tracking, 3D free form object, Skeletonization, Graph matching.

Incremental trajectory aggregation in video sequences

This article introduces new similarity measures between trajectories, in order to detect uncommon behaviors. These measures are used to find the most common trajectories in a sequence, using an implicit aggregation method. They may be applied to trajectories of objects tracked in real time. Moreover, by combining one or more measures, it is possible to variate the impact of the temporal dimension - velocity along a trajectory. Our experiments show that the measures are able to properly identify rare trajectories in a video, as well as to detect the most frequent ones.

Real-time Multi-Object Tracking with Occlusion and Stationary Objects Handling for Conveying Systems

Multiple object tracking has a broad range of applications ranging from video surveillance to robotics. In this work, we extend the application field to automated conveying systems. Inspired by tracking methods applied to video surveillance, we follow an on-line tracking-by-detection approach based on background subtraction. The logistics applications turn out to be a challenging scenario for existing methods. This challenge is twofold: First, conveyed objects tend to have a similar appearance, which makes the occlusion handling difficult. Second, they are often stationary, which make them hard to detect with background subtraction techniques. This work aims to improve the occlusion handling by using the order of the conveyed objects. Besides, to handle stationary objects, we propose a feedback loop from tracking to detection. Finally, we provide an evaluation of the proposed method on a real-world video.

PaTHOS: Part-Based Tree Hierarchy for Object Segmentation

The problem we address in this paper is the segmentation and hierarchical grouping in digital images. In terms of image acquisition protocol, no constraints are posed to the user. At first, a histogram thresholding provides numerous segments where a homogeneity criterion is respected. Segments are merged together using similarity properties and aggregated in a hierarchy based on spatial inclusions. Shape and color features are extracted on the produced segments. Tests performed on Oxford Flower 17 [8] show that our method outperforms a similar one and allow the relevant object selection from the hierarchy. In our case, this approach represents the first stage towards flower variety identification.

Multiple Object Tracking on Smart Embedded Camera For Automated Conveying Systems

In this paper we present an embedded smart cameras system that performs multiple objects tracking for the control of an automated conveying system. The smart cameras automatically extract scene description models of the conveying environment from underlying motion structure, as observed over time. In particular, the ROI mask and the conveyors pathways model are used to improve the tracking performance during on-line inference. An efficient parallelization strategy of background subtraction is proposed. Our approach takes advantage of multi-core embedded platforms. We provide an evaluation of the proposed method on videos coming from an industrial application.