Hans-Hellmut Nagel

Displacement vectors derived from second-order intensity variations in image sequences

A local approach for interframe displacement estimates is developed by minimization of the squared differences between a second-order Taylor expansion of gray values from one frame and the observed gray values within the same window from the next frame. If the second-order terms in the Taylor expansion are significant, a system of two coupled nonlinear equations for the two unknown components of the displacement vector can be derived. In the special case of “gray value corners,” these equations can be simplified to facilitate a closed form solution. An iterative refinement procedure is developed to extend these estimates for image regions which do not exhibit exactly the properties of “gray value corners.” The minimization approach is generalized in such a way that the approach of Horn and Schanck (Artif. Intell. 17, 1981, 185–203) can be recognized as a special case of this generalized form which should be applicable even across occluding edges. It thus appears to be an interesting model for the local computation of optical flow.

On the Analysis of Accumulative Difference Pictures from Image Sequences of Real World Scenes

The count of events where sample areas from the second and subsequent frames of a TV-image sequence are incompatible with the corresponding sample area of the first frame are accumulated in a first-order difference picture (FODP). Analysis of this FODP provides a separate estimate for images of moving objects and of stationary scene components. We start from the hypothesis that the first frame represents the stationary scene component. Once it has been recognized that a subarea of this initial estimate corresponds to the image of a moving object, the grey values in this subarea are replaced by later estimates of the stationary background at this position. No knowledge specific to a particular scene is utilized in the algorithm. The results for two scene sequences are presented.

Volumetric model and 3D trajectory of a moving car derived from monocular TV frame sequences of a street scene

A polyhedral approximation for the volumetric description of a moving rigid object from a real-world scene is derived, based on measurements in monocular TV-frame sequences. The trajectory and attitude of the object motion relative to the camera is simultaneously determined up to the same factor which scales the object description. Results from one street scene sequence are presented. The approach is compared to related ones reported in the recent literature.

New likelihood test methods for change detection in image sequences

Abstract Modeling the image as a piecewise linear gray-value function of the pixel coordinates considerably improved a change detection test based previously on a piecewise constant gray-value function. These results encouraged investigations into modeling the picture as a mosaic of patches where the gray-value function within each patch is described as a second-order bivariate polynomial of the pixel coordinates. Such a more appropriate model allowed the assumption to be made that the remaining gray-value variation within each patch can be attributed to noise related to the sensing and digitizing devices, independent of the individual image frames in a sequence. This assumption made it possible to relate the likelihood test for change detection to well-known statistical tests ( t test, F test), facilitating the determination of threshold values related to a priori confidence limits.

Formation of an object concept by analysis of systematic time variations in the optically perceptible environment

Abstract An algorithm to isolate a representation for a moving object of unknown form and size from a sequence of TV frames is described, together with results of its application to real world scenes. It is based on the assumption that a group of connected regions being jointly displaced in a systematic way from frame to frame without changing their relative positions represents the image of a moving object. These assumptions relate to properties which are to a large extent independent of specific scenes and thus may appear as components in a general object concept. Implications of this approach for the conception of scene analyzing systems that “learn from observation” are discussed.

Overview on Image Sequence Analysis

Attempts to concisely describe dynamic phenomena recorded by image sequences tend to model the depicted scene as a configuration of objects which exhibit well defined (sequences of) state transitions. Current research in image sequence analysis is concerned with the design of computer-internal representations for objects and associated dynamic phenomena.

On change detection and displacement vector estimation in image sequences

Recent developments in the areas of displacement vector estimation as well as dissimilarity grading by a maximum likelihood ratio can be related to each other quantitatively in such a way that dissimilarity grading is reduced to interframe displacement estimation.

On the Frame-to-Frame Correspondence between Grey Value Characteristics in the Images of Moving Objects

A system approach is outlined for the derivation of 3D polyhedral descriptions of moving objects by evaluation of monocular TV-frame sequences from real-world scenes. An implementation of this approach facilitated the study of the correspondence problem between descriptors extracted from images of moving cars in consecutive TV-frames. Our experience forced us to modify the relaxation approach of Barnard and Thompson 79+80 [1] in order to obtain acceptable results. These modifications are described and discussed.

Postlude to a PASCAL-compiler bootstrap on a decsystem-10

There can be a considerable difference between a portable compiler and a usable one which is a high level language implementation with a fair chance of acceptance by the general public in competition with other—vendor supported—systems. The experiences gained in pushing a portable PASCAL compiler across such an acceptance barrier are discussed in detail. Comments on some shortcomings of the PASCAL definition and suggested remedies are included.

Ueber die Repraesentation von Wissen zur Auswertung von Bildern

Als Rahmen fuer die Diskussion von Segmentationsalgorithmen hat Kanade 78 ein Modell der algorithmischen Bilddeutung (image understanding) vorgelegt. Dieses Modell bietet eine Struktur von Konzepten an, mit deren Hilfe nicht nur Segmentations-Algorithmen sondern auch darueber hinausgehende Fragen der algorithmischen Bilddeutung behandelt werden koennen. Aufbauend auf einer kurzen Darlegung dieses Modells von Kanade sollen eine Reihe von Problemen bei der Repraesentation von Wissen zur Auswertung von Bildern diskutiert werden.