Some of today's safety and driver assistance systems are mainly based on either radar or camera systems. The fusion of both sensors will generate synergies in Adaptive Safety Systems. Focus of the presentation will be on the current sensor fusion approaches.

The increased demand for more intelligence in automotive applications requires robust low-cost sensory systems. This paper presents low-cost high-speed camera employing novel CMOS sensors capable of serving in different automotive applications. Our camera offers a platform that can be used to perform different tasks like occupancy detection, precrash sensing, collision avoidance, and parking distance control for the driver assistance. Moreover, high-speed camera can be used in component development e.g. in the analysis of high-speed motion in crash-test.

Acquisition of the images of fast-moving objects requires imagers with high photoresponsivity at short integration times, synchronous exposure, and high-speed parallel readout. Previous CMOS implementations yield frame rates around 500 frames/s at integration times ranging from 75 to 200 ps, and some use rolling shutter only. This CMOS imager achieves more than 1000 frames/s with integration time in synchronous exposure variable between 1 /spl mu/s and 150 /spl mu/s. The 256/spl times/256 pixel imager is realized in a single-poly double-metal n-well 1 /spl mu/m CMOS technology.

In this paper a new camera system for high speed imaging is presented which is capable of recording images with a resolution of 256x256 pixels and frame rates in excess of 1000 frames per second. It uses an image sensor with on chip electronic shutter and has been fabricated in standard 1 mum standard CMOS (Complementary Metal Oxide Semiconductor) process. The camera system contains an image memory, for sequence recording. The camera delivers a very, good image qualify without any external algorithm for image enhancement and provides a very fast interface between the image acquisition and image processing unit. Another among several advantages of CMOS imagers compared to their CCD (Charged Coupled Device) counterparts is the flexibility and the possibility, to acquire images in a very short period. This allows an adaptation of the camera to various automotive applications like occupancy detection airbag control, precrash sensing, collision avoidance, surveillance, and crash test observation. Moreover, the sl stem architecture makes a combination of several applications possible using just a single image sensor unit.

This paper presents an unsupervised texture segmentation algorithm based on feature extraction using multichannel Gabor filtering. It is shown that feature contrast, a criterion derived for Gabor filter parameter selection, is well suited for feature coordinate weighting in order to reduce the feature space dimension. The central idea of the proposed segmentation algorithm is to decompose the actual segmented image into disjunct areas called scrap images and use them after lowpass filtering as additional features for repeated k-means clustering and minimum distance classification. This yields a classification of texture regions with an improved degree of homogeneity while preserving precise texture boundaries

The present paper describes an automated procedure for the detection of left ventricular internal wall edges in digital echocardiographic image sequences. The proposed procedure is divided into three steps and programmed in C/UNIX. It includes the use of a specially designed, application-specific adaptive spatio-temporal filter for noise reduction in echocardiographic image sequences, local 3-D histogram equalization for augmented contrast, and segmentation of the left ventricle using a regional growth method. When designing the adaptive spatio-temporal filter, we took into account the fact that the background noise is correlated in tangential orientation due to beam deflection at interfaces characterized by a large impedance ,,jump. Using the specially designed filter, the background noise is successfully reduced without degrading the ventricular contours. The simulation results presented highlight the performance of the proposed method in an exemplary manner.

The aim of motion detection in image sequences for sceneanalysis is long established. Also, it can be used inapplications like surveillance or occupant detection forsegmentation of moving objects as the first signalprocessing step. This contribution presents a novel imageprocessing system based on a CMOS image sensor installedat the car roof for many purposes such as interiorcompartment monitoring for theft prevention or objectrecognition. The presented approach concentrates on...

The present paper describes an automated procedure for the detection of left ventricular internal wall edges in digital echocardiographic image sequences. The proposed procedure is divided into three steps and programmed in C/UNIX. It includes the use of a specially designed, application-specific adaptive spatio-temporal filter for noise reduction in echocardiographic image sequences, local 3-D histogram equalization for augmented contrast, and segmentation of the left ventricle using a regional growth method. When designing the adaptive spatio-temporal filter, we took into account the fact that the background noise is correlated in tangential orientation due to beam deflection at interfaces characterized by a large impedance ('')jump". Using the specially designed filter, the background noise is successfully reduced without degrading the ventricular contours. The simulation results presented highlight the performance of the proposed method in an exemplary manner.

The present paper describes an automated procedure for the detection of left ventricular internal wall edges in digital echocardiographic image sequences. The proposed procedure is divided into three steps and programmed in C/UNIX. It includes the use of a specially designed, application-specific adaptive spatio-temporal filter for noise reduction in echocardiographic image sequences, local 3-D histogram equalization for augmented contrast, and segmentation of the left ventricle using a regional growth method. When designing the adaptive spatio-temporal filter, we took into account the fact that the background noise is correlated in tangential orientation due to beam deflection at interfaces characterized by a large impedance "jump". Using the specially designed filter, the background noise is successfully reduced without degrading the ventricular contours. The simulation results presented highlight the performance of the proposed method in an exemplary manner.

This article describes a novel approach to orientation and scale-invariant detection of textured objects in images. It performs both, a segmentation of multi-object scenes and the identification of rotation angles and scale rates of textures in an image by applying a comparison with reference texture features stored in a database. The main novelty of the proposed method is the transform of rotation and dilation into shifts in the feature space by employing a polar-log Gabor filter bank. Texture segmentation and identification of the rotation angles and scale rates have been carried out using symmetric phase only matched filters. The simulation results illustrated highlight the performance of the presented method in an exemplary manner.

In recent studies on image analysis an increasing effort has been carried out in the area of wavelet transform techniques for the discrimination and classification of textural images. These methods compete with multichannel filtering techniques, especially with the nonorthogonal and incomplete Gabor filtering. In this paper we introduce two feature extraction algorithms based on pyramidal and tree structured wavelet transforms and compare their performance with the feature extraction which employs adaptive Gabor filtering. This comparison is based on the segmentation results of several texture image examples using the identical segmentation algorithm for all three feature extraction methods. The visible differences of the segmentation results are discussed and their algorithmic causes are analysed.

The present study describes an automatic procedure for calculating the ejection fraction on the basis of ultrasound images. This procedure is divided into three steps, and programmed in C/UNIX. In an initial step, an algorithm, known as sequence analysis, is used to identify the enddiastolic and endsystolic images from a series of ultrasound images. With the use of a template matching procedure for calculating displacement vectors, a method has been developed for the first time for analysing moving images for identification purposes. The subsequent image analysis employs various filter processes for noise suppression, contrast enhancement, image sharpening, edge detection and segmentation. On the basis of the geometrical data of the segmented left ventricle obtained in this way the respective volumes and ejection fraction are calculated. To validate this automatic procedure, the series of ultrasound images were examined by a cardiologist. It was found that the identification of the endsystolic and enddiastolic image was correct, and that there was good agreement between the automatically determined and manually determined ejection fraction.

The present study describes an automatic procedure for calculating the ejection fraction on the basis of ultrasound images. This procedure is divided into three steps, and programmed in C/UNIX. In an initial step, an algorithm, known as sequence analysis, is used to identify the enddiastolic and endsystolic images from a series of ultrasound images. With the use of a template matching procedure for calculating displacement vectors, a method has been developed for the first time for analysing moving images for identification purposes. The subsequent image analysis employs various filter processes for noise suppression, contrast enhancement, image sharpening, edge detection and segmentation. On the basis of the geometrical data of the segmented left ventricle obtained in this way the respective volumes and ejection fraction are calculated. To validate this automatic procedure, the series of ultrasound images were examined by a cardiologist. It was found that the identification of the endsystolic and enddiastolic image was correct, and that there was good agreement between the automatically determined and manually determined ejection fraction.

We present a segmentation algorithm for multichannel image analysis. It is based on a novel method that significantly improves the segmentation performance with respect to both homogeneity of the segmented regions and precision of the segmented region boundaries. The algorithm yields excellent results in comparison with other segmentation algorithms that are based on feature space clustering followed by minimum distance classification, as is shown in some segmentation examples. The main idea of the algorithm is the iterative feedback of the knowledge about the analysed image that has been obtained from preceding segmentation results. It needs just a stack of feature images and the indication of the number of required classes for input data. Therefore, it has a broad field of possible applications, especially in multichannel image analysis.