Michael Hötter

Object-oriented analysis-synthesis coding of moving images

Abstract An object-oriented analysis-synthesis coder is presented which encodes objects instead of blocks of N × N picture elements. The objects are described by three parameter sets defining the motion, shape and colour of an object. The parameter sets are obtained by image analysis based on source models of either moving 2D-objects or moving 3D-objects. Known coding techniques are used to encode the parameter sets. An object-depending parameter coding allows to introduce geometrical distortions instead of quantization errors. Using the transmitted parameter sets an image can be reconstructed by model-based image synthesis. Experimental results achieved with a first implementation of the coder are given and are discussed.

Object-oriented analysis-synthesis coding based on moving two-dimensional objects

Abstract An object-oriented analysis-synthesis coder is presented which encodes arbitrarily shaped objects instead of rectangular blocks. The objects are described by three parameter sets defining their motion, shape and colour. Throughout this contribution, the colour parameters denote the luminance and chrominance values of the object surface. The parameter sets of each object are obtained by image analysis based on source models of moving 2D-objects and coded by an object-dependent parameter coding. Using the coded parameter sets an image can be reconstructed by model-based image synthesis. In order to cut down the generated bit-rate of the parameter coding, the colour updating of an object is suppressed if the modelling of the object by the source model is sufficiently exact, i.e., if only a relatively small colour update information would be needed for an errorless image synthesis. Omitting colour update information, small position errors of objects denoted as geometrical distortions are allowed for image synthesis instead of quantization error distortions. Tolerating geometrical distortions, the image area to be updated by colour coding can be decreased to 4% of the image size without introducing annoying distortions. As motion and shape parameters can efficiently be coded, about 1 bit per pel remains for colour updating in a 64 kbit/s coder compared to about 0.1 bit per pel in the standard reference coder (RM8) of the CCITT. Experimental results concerning the efficient coding of motion and shape parameters are given and discussed. The coding of the colour information will be dealt with in further research.

A new system for video-based detection of moving objects and its integration into digital networks

In many video surveillance applications outdoor scenes are to be observed, normally located far away from the location of the observation personnel. This application scenario yields some basic demands if video detection techniques are applied: (a) The detection scheme has to be robust against distortions like varying illumination conditions, small camera motion, trees in motion, rain, snow, etc. (b) The video images and/or detection results have to be transmitted to the surveillance center. (c) The calibration and installation effort of the sensor should be as simple as possible. In conventional schemes, an individual transmission channel is necessary for each sensor connected to a surveillance center. The costs increase with the distance and the number of connections. As a video signal is transmitted the connection needs high bandwidth. In this paper, a new system is presented which includes a video-based detection of moving objects in natural scenes and a transmission of images via digital networks. The detection and description of moving objects is based on an object-oriented, statistical multi-feature analysis of video sequences. This analysis is self-adapting to an observed scene, such that the calibration effort is very low. In case of an alarm event, object parameters are extracted and video images are memorized showing the history of the alarm event. Applying compression techniques this data are transmitted to an arbitrarily located surveillance center using digital networks

Detection and description of moving objects by stochastic modelling and analysis of complex scenes

Abstract This paper presents a new technique for the detection and description of moving objects in natural scenes which is based on a statistical multi-feature analysis of video sequences. In most conventional schemes for the detection of moving objects, temporal differences of subsequent images from a video sequence are evaluated by so-called change detection algorithms. These methods are based on the assumption that significant temporal changes of an image signal are caused by moving objects in the scene. However, as temporal changes of an image signal can as well be caused by many other sources (camera noise, varying illumination, small camera motion), such systems are afflicted with the dilemma of either causing many false alarms or failing to detect relevant events. To cope with this problem, the additional features of texture and motion beyond temporal signal differences are extracted and evaluated in the new algorithm. The adaptation of this method to normal fluctuations of the observed scene is performed by a time-recursive space-variant estimation of the temporal probability distributions of the different features (signal difference, texture and motion). Feature data which differ significantly from the estimated distributions are interpreted to be caused by moving objects.

Investigations on colour coding in an object-oriented analysis: synthesis coder

Abstract In this paper, a new colour coding algorithm called Hybrid Adaptive DCT/DPCM Colour Coding is presented which encodes the colour parameters of objects in an object-oriented analysis-synthesis coder with a hybrid scheme, where either a DPCM (Differential Pulse Code Modulation) technique or a DCT (Discrete Cosine Transform) is used whichever allows a more efficient coding. In experimental results, the coding efficiency of the Hybrid Adaptive DCT/DPCM Colour Coding is compared to purely block-oriented DCT coding and region-oriented transform coding for typical videophone sequences at data rates of about 64 kbit/s. The relative gain concerning the average bit-rate at a fixed image quality is about 5% compared to region-oriented transform coding and 41% compared to block-oriented DCT coding. Beside its coding efficiency, Hybrid Adaptive DCT/DPCM Coding can easily be realized by fast algorithms of low computational complexity.

Detection of moving objects in natural scenes by a stochastic multi-feature analysis of video sequences

A new technique for the detection and description of moving objects in natural scenes is presented which is based on an object-oriented, statistical multi-feature analysis of video sequences. In most conventional schemes for the detection of moving objects, temporal differences of subsequent images from a video sequence are evaluated in a block based manner by so called change detection algorithms. These methods are based on the assumption that significant temporal changes of an image signal are caused by moving objects in the scene. However, as temporal changes of an image signal can as well be caused by many other sources (camera noise, varying illumination, small camera motion, trees in motion), such systems are afflicted with the dilemma of either causing many false alarms or failing to detect relevant events. To scope with this problem, the additional features texture and motion beyond temporal signal differences are extracted and evaluated in the new algorithm. Furthermore, these features are evaluated in an object-oriented instead of a block oriented fashion to increase the reliability of detection. The adaption of this method to normal fluctuations of the observed scene is performed by a time-recursive space-variant estimation of the temporal probability distributions of the different features (signal difference, texture and motion). Feature data which differ significantly from the estimated distributions are interpreted to be caused by moving objects.

Detection of moving objects using a robust displacement estimation including a statistical error analysis

A new technique for the detection and description of moving objects in natural scenes is presented which is based on an object-oriented, statistical multifeature analysis of video sequences. To cope with the problem that image signal changes can have causes other than object motion, additional features, viz, texture and motion beyond temporal signal differences, are extracted and evaluated in an object-oriented fashion. The adaptation of this method to normal fluctuations of the observed scene is performed by a time-recursive space-variant estimation of the temporal probability distributions of the features. Feature data which differ significantly from the estimated distributions are interpreted to be caused by moving objects. For motion feature extraction, a robust displacement estimation algorithm is applied which is oriented towards the joint estimation of displacement vectors and their corresponding reliability measures. The reliability measures judge object motion and control the alarm setting. The advantages of the presented object detection algorithm compared to mere change detection techniques are demonstrated by some experiments. Due to its capability to automatically learn the observed scene the calibration effort of the sensor is extremely small.

Probabilistic evaluation of three-dimensional reconstructions from X-ray images spanning a limited angle.

An important part of computed tomography is the calculation of a three-dimensional reconstruction of an object from series of X-ray images. Unfortunately, some applications do not provide sufficient X-ray images. Then, the reconstructed objects no longer truly represent the original. Inside of the volumes, the accuracy seems to vary unpredictably. In this paper, we introduce a novel method to evaluate any reconstruction, voxel by voxel. The evaluation is based on a sophisticated probabilistic handling of the measured X-rays, as well as the inclusion of a priori knowledge about the materials that the object receiving the X-ray examination consists of. For each voxel, the proposed method outputs a numerical value that represents the probability of existence of a predefined material at the position of the voxel while doing X-ray. Such a probabilistic quality measure was lacking so far. In our experiment, false reconstructed areas get detected by their low probability. In exact reconstructed areas, a high probability predominates. Receiver Operating Characteristics not only confirm the reliability of our quality measure but also demonstrate that existing methods are less suitable for evaluating a reconstruction.

Moving Object Detection in Image Sequences Using Texture Features

Publisher Summary This chapter presents an algorithm for detection of moving objects in image sequences. The proposed algorithm uses texture features which are obtained blockwise from the frames of the image sequence. The object detection itself is essentially a temporal change detection algorithm, detecting changes of corresponding texture features between successive frames. The subsequent change detection algorithm operates on a small number of simple features computed per block. This approach results in an efficient object detection scheme with low computational complexity. The method is insensitive to small movements of strongly textured areas like trees moving in the wind. An object entering or leaving a block will, however, cause a change of the feature in almost all cases. Thus the reliability of the object detection can be increased by using suitable texture features. Sharing the low complexity with earlier detection methods, the presented algorithm can deal with complex textured scenes and temporarily varying image signal statistics. The robustness and efficiency of the proposed method has been extensively tested in offline simulations as well as in online processing of numerous scenes.

Zuverlässigkeit und Effizienz von Verfahren zur Verschiebungsvektorschätzung

Die Schatzung von Verschiebungsvektoren ist sowohl fur die Analyse von zeitlichen Bildsequenzen als auch fur die Interpretation von Stereobildpaaren von zentraler Bedeutung. Im Vordergrund dieses Beitrags steht die Frage, wie mit vertretbarem Aufwand eine moglichst zuverlassige Bestimmung der Verschiebungsvektoren moglich ist. Dabei wird berucksichtigt, das eine solche Messung nur in bestimmten Bildbereichen uberhaupt moglich ist, und das die resultierenden Meswerte stets im nachhinein auf ihre Zuverlassigkeit untersucht werden sollten. Das hier vorgeschlagene grundsatzliche Verarbeitungsschema lautet wie folgt: 1. Vorauswahl von Bildbereichen, in denen Verschiebungsvektoren ermittelt werden konnen 2. Berechnung der Verschiebungsvektoren in den ausgewahlten Bereichen 3. A-posteriori Bewertung der Zuverlassigkeit der ermittelten Verschiebungsvektoren