This paper presents a sensitivity analysis of using instantaneous multichannel two-dimensional (2D) imaging to achieve instantaneous 2D imaging spectroscopy. A simulated multiple-filter mosaic was introduced and used to acquire multichannel data which were transformed into spectra. The feasibility of two different transformation approaches (the concrete pseudoinverse approach and a statistical approach) was investigated through extensive experimental tasks. A promising statistical method was identified to be used for accurate estimation of spectra from multichannel data. Comparison between estimated and measured spectra shows that higher estimation accuracy can be achieved when using a larger number of usable multiple-filter combinations in the mosaic.

In this paper we present a new method for vision-based, reactive robot navigation that enables a robot to move in the middle of the free space by exploiting both central and peripheral vision. The robot employs a forward-looking camera for central vision and two side-looking cameras for sensing the periphery of its visual field. The developed method combines the information acquired by this trinocular vision system and produces low-level motor commands that keep the robot in the middle of the free space. The approach follows the purposive vision paradigm in the sense that vision is not studied in isolation but in the context of the behaviors that the system is engaged as well as the environment and the robot's motor capabilities. It is demonstrated that by taking into account these issues, vision processing can be drastically simplified still giving rise to quite complex behaviors. The proposed method does not make strict assumptions about the environment, requires very low level information to be extracted from the images, produces a robust robot behavior and is computationally efficient. Results obtained by bath simulations and from a prototype on-line implementation demonstrate the effectiveness of the method.

The authors classify edges with respect to diffuseness and profile type by using a tool evolved from scale space methodology called signatures. The method also enables the user to measure degree of diffuseness, height of edges, and bar edge width. The Gaussian difference signatures and apparent width signatures used are defined. To sift out the major edges, subsets of edge pixels must be selected. For describing the local gray level, landscape averaging needs to be done in a noncommittal fashion, as opposed to fitting a specific function blindly to data. Edge focusing, which is a coarse-to-fine traversal of brightness changes, deals with the subsets, whereas signatures may be thought of as a kind of reasonably cleverly done averaging. This method is of importance for distinguishing between illumination phenomena and object edges, performing depth-from-focus in single images, and in junction analysis.>

The author shows that there are two basic ways to eliminate rotation; either you eliminate actual (effective) rotation R or abstract rotation R. The transformation of visual directions P is decomposed into pure deformation (pure strain) and abstract rotation. Methods can be grouped as to whether they eliminate actual or abstract rotation. An interesting elimination of R can be done for equatorial motion fields on the image hemisphere, using equator-normal flow, having a moving sensor in a fixed environment. Then rotation can be eliminated by detecting the phase and amplitude of a sine wave hidden in a 1-D signal (the signal is a function of longitudes). Both sparse velocities and normal flow for nearly equator-parallel image contours can be used (directly) as input data. This approach seems robust and can be used in conjunction with the algorithm by R.C. Nelson and J. Aloimonos (1988).>

A method is developed for the measurement of short-range visual motion in image sequences, making use of the motion of image features such as edges and points. Each feature generates a Gaussian activation profile in a spatiotemporal neighborhood of specified scale around the feature itself; this profile is then convected with motion of the feature. The authors show that image velocity estimates can be obtained from such dynamic activation profiles using a modification of familiar gradient techniques. The resulting estimators can be formulated in terms of simple ratios of spatiotemporal filters (i.e. receptive fields) convolved with image feature maps. A family of activation profiles of varying scale must be utilized to cover a range of possible image velocities. They suggest a characteristic speed normalization of the estimate obtained from each filter in order to decide which estimate is to be accepted. They formulate the velocity estimators for dynamic edges in 1-D and 2-D image sequences, as well as that for dynamic feature points in 2-D image sequences.>

We analyze curves in motion. First, assume that we do not know the motion of points in the image plane, i.e. the optical flow. Then, define “ambiguous curves” as those that can give us nothing but ambiguous information about the underlying motion and - hence - the optical flow. The importance of a catalogue of am.bzguoas ciirues lies in telling us under what circumstances optical flow, motion and structure parameters cannot be extracted from moving curves (without point-to-point correspondence), irrespective of the method used. We give a catalogue of ambiguous curves , specializing to a planar patch in motion. Ambiguous curves (excluding the case of a single line) correspond either to a one- or a two-parameter family of solution coefficients, so the dimensionality of the non-unique solution can be predicted. Second, we wish to find analytical descriptions of the optical flow arising from objects in rigid motion, or, deforming linearly. M‘e stress the importance of using visual directions for optical flow descriptions ~ instead of image points. Modelling visual directions as elastic strings attached to a moving plane, we get a unique 3 x 3 matrix P describing the incremental transformation of visual directions. This description is more general than image point descriptions, and a convenient transform giving us concrete geometric interpretations when dealing with complicated dynamics. Using P, we can link critical-point analysis to the analysis of eigenvectors of P. The projected field lines for the optical flow, for a moving planar patch, can easily be derived from P. The two subjects, ambiguous curves and field lines of the optical flou are actually two sides of the same coin.

Edge detection in a gray-scale image at a fine resolution typically yields noise and unnecessary detail, whereas edge detection at a coarse resolution distorts edge contours. We show that ``edge focusing'', i.e., a coarse-to-fine tracking in a continuous manner, combines high positional accuracy with good noise-reduction. This is of vital interest in several applications. Junctions of different kinds are in this way restored with high precision, which is a basic requirement when performing (projective) geometric analysis of an image for the purpose of restoring the three-dimensional scene. Segmentation of a scene using geometric clues like parallelism, etc., is also facilitated by the algorithm, since unnecessary detail has been filtered away. There are indications that an extension of the focusing algorithm can classify edges, to some extent, into the categories diffuse and nondiffuse (for example diffuse illumination edges). The edge focusing algorithm contains two parameters, namely the coarseness of the resolution in the blurred image from where we start the focusing procedure, and a threshold on the gradient magnitude at this coarse level. The latter parameter seems less critical for the behavior of the algorithm and is not present in the focusing part, i.e., at finer resolutions. The step length of the scale parameter in the focusing scheme has been chosen so that edge elements do not move more than one pixel per focusing step.