An image-interpolation technique for the computation of optic flow and egomotion

 A technique for measuring the motion of a rigid, textured plane in the frontoparallel plane is developed and tested on synthetic and real image sequences. The parameters of motion – translation in two dimensions, and rotation about a previously unspecified axis perpendicular to the plane – are computed by a single-stage, non-iterative process which interpolates the position of the moving image with respect to a set of reference images. The method can be extended to measure additional parameters of motion, such as expansion or shear. Advantages of the technique are that it does not require tracking of features, measurement of local image velocities or computation of high-order spatial or temporal derivatives of the image. The technique is robust to noise, and it offers a simple, novel way of tackling the ‘aperture’ problem. An application to the computation of robot egomotion is also described. Received: 3 September 1993/Accepted in revised form: 16 April 1994     

Visual homing using an associative memory

Homing is the process by which an autonomous system guides itself to a particular location on the basis of sensory input. In this paper, a method of visual homing using an associative memory based on a simple pattern classifier is described. Homing is accomplished without the use of an explicit world model by utilizing direct associations between learned visual patterns and system motor commands. The method is analyzed in terms of a pattern space and conditions obtained that allow the system performance to be predicted on the basis of statistical measurements on the environment. Results of experiments utilizing the method to guide a robot-mounted camera in a three-dimensional environment are presented.     

Noninvasive prediction of vertebral body compressive strength using nonlinear finite element method and an image based technique

Noninvasive prediction of vertebral body strength under compressive loading condition is a valuable tool for the assessment of clinical fractures. This paper presents an effective specimen-specific approach for noninvasive prediction of human vertebral strength using a nonlinear finite element (FE) model and an image based parameter based on the quantitative computed tomography (QCT). Nine thoracolumbar vertebrae excised from three cadavers with an average age of 42 years old were used as the samples. The samples were scanned using the QCT. Then, a segmentation technique was performed on each QCT sectional image. The segmented images were then converted into three-dimensional FE models for linear and nonlinear analyses. A new material model was implemented in our nonlinear model being more compatible with real mechanical behavior of trabecular bone. A new image based MOS (Mechanic of Solids) parameter named minimum sectional strength ((σ_uA)_min) was used for the ultimate compressive strength prediction. Subsequently, the samples were destructively tested under uniaxial compression and their experimental ultimate compressive strengths were obtained. Results indicated that our new implemented FE model can predict ultimate compressive strength of human vertebra with a correlation coefficient (R² = 0.94) better than usual linear and nonlinear FE models (R² = 0.83 and 0.85 respectively). The image based parameter introduced in this study ((σ_uA)_min) was also correlated well with the experimental results (R² = 0.86). Although nonlinear FE method with new implemented material model predicts compressive strength better than the (σ_uA)_min, this parameter is clinically more feasible due to its simplicity and lower computational costs. This can make future applications of the (σ_uA)_min more justified for human vertebral body compressive strength prediction.     

Recursive implementations of temporal filters for image motion computation

 Efficient algorithms for image motion computation are important for computer vision applications and the modelling of biological vision systems. Intensity-based image motion computation proceeds in two stages: the convolution of linear spatiotemporal filter kernels with the image sequence, followed by the non-linear combination of the filter outputs. If the spatiotemporal extent of the filter kernels is large, then the convolution stage can be very intensive computationally. One effective means of reducing the storage required and computation involved in implementing the temporal convolutions is the introduction of recursive filtering. Non-recursive methods require the number of frames of the image sequence stored at any given time to be equal to the temporal extent of the slowest temporal filter. In contrast, recursive methods encode recent stimulus history implicitly in the values of a small number of variables updated through a series of feedback equations. Recursive filtering reduces the number of values stored in memory during convolution and the number of mathematical operations involved in computing the filters' outputs. This paper extends previous recursive implementations of gradient- and correlation-based motion analysis algorithms [Fleet DJ, Langley K (1995) IEEE PAMI 17: 61–67; Clifford CWG, Ibbotson MR, Langley K (1997) Vis Neurosci 14: 741–749], describing a recursive implementation of causal band-pass temporal filters suitable for use in energy- and phase-based algorithms for image motion computation. It is shown that the filters' temporal frequency tuning curves fit psychophysical estimates of the temporal properties of human visual filters [Hess RF, Snowden RJ (1992) Vision Res 32: 47–60]. Received: 20 April 1999 /Accepted in revised form: 8 November 1999     

The representation of egomotion in the human brain

An essential function of visual processing is to establish the position of the body in space and, in concert with the other sense systems, to monitor movement of the whole body, or "egomotion." A key cue to egomotion is optic flow. For example, forward motion through the environment generates an expanding pattern of flow on the retina, and (with eyes fixed centrally) the direction of heading corresponds to the center of expansion [1]. In macaques, visual cortical area MST is sensitive to optic-flow structure [2, 3], and it has been suggested that MST has a central role in the computation of heading [4]. However, here we identify two areas of the human brain that represent visual cues to egomotion more directly than does MST. These areas respond strongly to a single optic-flow stimulus but become relatively unresponsive when the stimulus is surrounded with further flow patches and thereby made inconsistent with egomotion. One is putative area VIP in the anterior portion of the intraparietal sulcus. The other is a new visual area, which we refer to as cingulate sulcus visual area (CSv). Areas V1-V4 and MT respond about equally to both types of flow stimulus. MST has intermediate properties, responding well to multiple patches but with a modest preference for a single, egomotion-compatible patch. We suggest that MST is merely an intermediate processing stage for visual cues to egomotion and that such cues are more comprehensively encoded by VIP and CSv.     

Visual computation and saccadic eye movements: a theoretical perspective

A simple instance of parallel computation in neural networks occurs when the eye orients to a novel visual target. Consideration of target-elicited saccadic eye movements opens the question of how spatial position is represented in the visual pathways involved in this response. It is argued that a point-for-point retinotopic coding of spatial position (the 'local sign' approach) is inadequate to account for the characteristics of the response. An alternative approach based on distributed coding is developed.     

Biological computation of image motion from flows over boundaries.

A theory of early motion processing in the human and primate visual system is presented which is based on the idea that spatio-temporal retinal image data is represented in primary visual cortex by a truncated 3D Taylor expansion that we refer to as a jet vector. This representation allows all the concepts of differential geometry to be applied to the analysis of visual information processing. We show in particular how the generalised Stokes theorem can be used to move from the calculation of derivatives of image brightness at a point to the calculation of image brightness differences on the boundary of a volume in space-time and how this can be generalised to apply to integrals of products of derivatives. We also provide novel interpretations of the roles of direction selective, bi-directional and pan-directional cells and of type I and type II cells in V5/MT.     

3-D reconstruction of biological objects using underwater video technique and image processing

This paper describes a 3-D reconstruction method which allows accurate measurements of volume, surface area and other morphometric measurements of three-dimensional biological objects, without removing them from the sea. It represents a novel approach based on multiple views (eight resulted to be sufficient) from underwater video images and a new image processing procedure (MOD3D), whose application has met the basic requirements (i.e. to work on images recorded in turbid waters, with nonuniform lighting, to investigate large areas and in reasonable time, etc.) imposed when operating in the marine environment with simple, easy-to-use and nonprofessional equipment. It is a noninvasive, nondestructive and in the field fast method, thus suitable for sampling also at relevant depth, whose applicability has specifically been set up for a range of growth forms from massive to submassive and irregularly shaped. The accuracy of the method was assessed using models with three levels of 3-D complexity: simple, moderate and complex morphology. A high accuracy of volume measurements made through MOD3D image analysis software was achieved when compared with the laboratory water displacement method, which represents the most accurate method for volume measurement, with an overall mean percent error of about 1.7% (S.D. 2.2%). For all three levels of morphologic complexity, no significant differences (p>0.05) were found. Volume measurements obtained in field based on geometric approximation resulted rough, with significant differences from the MOD3D values (p<0.05). The geometric approximation was lower than MOD3D for simple and moderate morphology, and variable for complex morphology. For all three models, MOD3D values for surface area computation were consistently lower (mean error 13%) than the foil-wrapping values (p<0.05), due to overlap error when foil wrapping. Two applications were made with the bryozoan Pentapora fascialis and the coral Cladocora caespitosa to quantify carbonate standing stock and biomass of these two carbonate framework builders, whose importance has been recently recognised among the temperate sublittoral benthic species. Time required for the 3-D reconstruction method (about 3 h) makes it suitable for routine application particularly for relatively large area investigations, with irregularly shaped objects on rough substrate and several biological objects within the area.     

Simplified measurement of soil pH using an agar-contact technique

A method for the indirect measurement of soil-pH is described. This method allows the spatial arrangement of soil and rhizosphere to be conserved. The soil is brought into contact with a layer of agar, containing bromocresol purple. A nylon gauze is placed between soil and agar. For quantitative pH measurements, a micro-electrode is inserted into the agar after three hours of contact between soil and agar.The validity of the method was checked by comparing its results with those obtained by standard procedures. At different pH-levels (pH 5.0 to 7.0) in either a sandy or a clay soil, a high correlation (r²=0.98) was found between the two methods. However, in the case of the clay soil, the agar-pH was significantly lower than the standard-pH. In the sandy soil, in the range pH 5.0 to 6.0, the results of both methods agreed very well. The agar method was used to measure the pH dynamics in the rhizosphere of lucerne seedlings, grown in rhizotrons.     

An active method of extracting egomotion parameters from optical flow

Instead of the former passive method characterized by a camera fixed on a translating and rotating vehicle, an active method characterized by tracking an object at the center of visual field is proposed. The method extracts egomotion parameters such as the instantaneous direction of translation, the axis of rotation and its angular velocity. A unit spherical surface is used as the projection surface.The theory allows clear vision where needed at the center of the visual field, and simultaneously permits the extraction of egomotion parameters from the periphery of the visual field.     

Classification and counting of fluorescent pollen using an image analysis system

Pollen viability is commonly assessed by fluorochromatic reaction (FCR) because of the high correlation between positive fluorescence of the pollen grains and their ability to germinate. One of the advantages of this method is its simplicity. An experiment to test FCR analysis for reproducibility, however, showed that results are affected by subjectivity. There is little consistency between analysts, and assessment by the same analyst may differ for the same pollen sample image examined at different times. These problems were solved by a computerized image analysis system that provides a method for classifying positive and negative fluorescent pollen and automatic counting of the grains in each class. The computerized image analysis system does not change the biochemistry of the FCR test, but avoids some experimental errors owing to the subjectivity of the analyst. Microscope images of the pollen after FCR were digitized and later analyzed by specially designed software, Plant Meter. This software deletes the dark background of the image to isolate the grains, and subsequently counts positive and negative fluorescent pollen grains. An experiment was carried out to validate software output and it showed reliable results. Moreover, the software is user friendly and very little training is necessary for analysts to achieve reliable results.     

Extended production of citric acid using an exchange filtration technique

A filtration technique has been applied to deep culture citric acid fermentation in an attempt to prolong the active phase of citric acid production. The fermentation was extended to 30 days with a daily production rate of between 9 and 10g^–1d^–1. A yield of over 70% citric acid was achieved.     

Parallel computation of simple arithmetic using peptide-antibody interactions

We propose a theoretical model for representing and manipulating binary numbers using peptide-antibody interactions. In particular, we present models to solve simple binary arithmetical problems like addition and subtraction. As the interactions can take place in parallel we show that the number of steps is independent of the size (bits) of the numbers.     

Estimating species trees using approximate Bayesian computation

Development of methods for estimating species trees from multilocus data is a current challenge in evolutionary biology. We propose a method for estimating the species tree topology and branch lengths using approximate Bayesian computation (ABC). The method takes as data a sample of observed rooted gene tree topologies, and then iterates through the following sequence of steps: First, a randomly selected species tree is used to compute the distribution of rooted gene tree topologies. This distribution is then compared to the observed gene topology frequencies, and if the fit between the observed and the predicted distributions is close enough, the proposed species tree is retained. Repeating this many times leads to a collection of retained species trees that are then used to form the estimate of the overall species tree. We test the performance of the method, which we call ST-ABC, using both simulated and empirical data. The simulation study examines both symmetric and asymmetric species trees over a range of branch lengths and sample sizes. The results from the simulation study show that the model performs very well, giving accurate estimates for both the topology and the branch lengths across the conditions studied, and that a sample size of 25 loci appears to be adequate for the method. Further, we apply the method to two empirical cases: a 4-taxon data set for primates and a 7-taxon data set for yeast. In both cases, we find that estimates obtained with ST-ABC agree with previous studies. The method provides efficient estimation of the species tree, and does not require sequence data, but rather the observed distribution of rooted gene topologies without branch lengths. Therefore, this method is a useful alternative to other currently available methods for species tree estimation.     

SCANGRAPH: a program for digitization of graphs using an image scanner

The program SCANGRAPH has been developed for vectorizing anydrawing or graph in the form of a single continuous line. Withthe use of the software, the raster image of a graph in theTIFF format as generated from an optical scanner or digitizingvideo system can be converted to the corresponding digitizedgraph with x and y coordinates given. The digitized data canbe output in the ASCII format for importing to other softwarepackages. Options for peak identification, data smoothing andcompression, and graph comparison are also provided for datamanipulation.