Object recognition and image understanding: theories of everything?
Explore the source record for details and available documents.
Computer Simulation↗
Biomedical subjects
T Caelli
Publications and source records attributed to T Caelli.
Learning paradigms for image interpretation.
In this paper we discuss image understanding and object recognition as a class of processes which involve binding what is seen with what is known. It follows from this perspective that it is important to explicate how systems may learn about spatial information from images, how it is encoded, and, how, in turn, such knowledge is matched with new image data. As well as discussing very specific solutions to these problems, this paper questions the view that our scientific understanding of image understanding is sufficiently circumscribed by the study of how image features are extracted and matched, per se.
Computer Simulation↗
Theory of spatiochromatic image encoding and feature extraction.
We consider how to interpret, filter, and cross-correlate complex-value color (hue and saturation) images by using a single discrete Fourier transform: the spatiochromatic discrete Fourier transform. The model defines new types of spatiochromatic oriented sinusoidal gratings, termed rainbow gratings, which encode the variation of color over space. We demonstrate how color-opponent detectors observed within the vertebrate visual system can be easily defined by linear filters within this representation. This model also allows us to filter and detect both spatial and chromatic patterns in images by using a single cross-correlation procedure. In doing so, we explore a new form of the Cauchy-Schwartz inequality applied to complex-valued scalar products. Results demonstrate the power of this form of spatiochromatic matched filtering in detecting signals embedded in such a significant amount of noise that they are not visible to the unaided human eye.
Animals↗
Evidence-Based Pattern Classification: A Structural Approach to Human Perceptual Learning and Generalization
Models of human pattern classification have been traditionally based on implicit pattern descriptions which involve lists of continuous attribute values or discrete features. Here we propose an alternative approach which makes explicit use of pattern structure in terms of components and their unary (part-specific) and binary (part-relational) properties. Such attributes "evidence" different classes of patterns and allow one to model processes of both perceptual learning and generalization to novel instances. An object in an evidence-based system is represented by a set of rules, where each rule provides a certain amount of class-specific evidence. The accumulated class evidence over all activated rules determines the classification probability. We have examined how well this concept reflects human performance by training observers to classify compound Gabor patterns and then testing them with segmented (grey-level-transformed) versions of the patterns in the original training set. If the observers were to construct rules to define each pattern class in terms of perceived parts and their relations, then it should be expected that classification performance would generalize to these new patterns. Results confirm this hypothesis and the specific feature extraction, learning, and rule generation model used to predict performance. Copyright 1997 Academic Press
Journal Article↗
The IPRS Image Processing and Pattern Recognition System.
IPRS is a freely available software system which consists of about 250 library functions in C, and a set of application programs. It is designed to run under UNIX and comes with full source code, system manual pages, and a comprehensive user's and programmer's guide. It is intended for use by researchers in human vision, pattern recognition, image processing, machine vision and machine learning.
Humans↗
Recognition-by-parts: a computational approach to human learning and generalization of shapes.
In this paper human pattern recognition is modeled in terms of how human observers learn to describe patterns in terms of their perceived parts, their unary (part) and binary (relational) attributes and the way in which such attribute states "evidence' different classes of shapes. This approach, originally developed in the area of computer vision, is concerned with algorithms which enable the learning of shape descriptions from examples and the classification of new data (generalization) efficiently and accurately. An object in such an "evidence-based' system is represented by a set of rules, where each rule provides a certain amount of evidence for each object class in the database. The accumulated class evidence over all activated rules can then be used to determine the classification probability. We have examined how well this model reflects human perception by training observers to classify compound Gabor patterns and then testing them with versions of such patterns which were segmented (gray-level transformed) versions of the original training set. If the observers were to construct rules to define each pattern class in terms of perceived parts and their relations, then it should be expected that classification performance would generalize to these new patterns from the original set. Results confirm this hypothesis and the specific feature extraction, learning and rule generation model used to predict performance.
Attention↗
Generalization of form in visual pattern classification.
Human observers were trained to criterion in classifying compound Gabor signals with symmetry relationships, and were then tested with each of 18 blob-only versions of the learning set. Generalization to dark-only and light-only blob versions of the learning signals, as well as to dark-and-light blob versions was found to be excellent, thus implying virtually perfect generalization of the ability to classify mirror-image signals. The hypothesis that the learning signals are internally represented in terms of a 'blob code' with explicit labelling of contrast polarities was tested by predicting observed generalization behaviour in terms of various types of signal representations (pixelwise, Laplacian pyramid, curvature pyramid, ON/OFF, local maxima of Laplacian and curvature operators) and a minimum-distance rule. Most representations could explain generalization for dark-only and light-only blob patterns but not for the high-thresholded versions thereof. This led to the proposal of a structure-oriented blob-code. Whether such a code could be used in conjunction with simple classifiers or should be transformed into a propositional scheme of representation operated upon by a rule-based classification process remains an open question.
Adult↗
Machine learning paradigms for pattern recognition and image understanding.
In this paper some issues are considered related to the encoding of spatial information and associated perceptual learning algorithms which, it is claimed, are necessary for robust pattern and object recognition in multi-object (natural) scenes. The types of learning requirements within a 'recognition-by-parts' paradigm are contrasted with findings from alternative models.
Form Perception↗
Probabilistic analysis of human supervised learning and classification.
Probabilistic classification techniques based on Bayesian decision theory are used to analyze human supervised learning and classification. The procedure rests on the assumption that human classification behaviour is based on internal feature states which can be linked to physical feature vectors (corresponding to the system input). In the present approach, this relationship is modeled in terms of additive stochastic error signals. The corresponding random variables describe the additional degrees of bias and variance introduced by the (perceptual) process of internal feature measurement. Estimates of internal feature states are obtained by least-squares minimization. Structure and dimensionality of the resulting internal representation are displayed by plotting the configuration of internal class means, or virtual prototypes. Their temporal evolution reflects the dynamic properties of the learning process. The use of the procedure is demonstrated by analyzing the results of two experiments. First, it is shown that Minimum Distance Classifiers, such as used by Caelli, Rentschler and Scheidler [(1987) Biological Cybernetics, 57, 233-240], are suboptimal in predicting human performance. Second, it is found that extrafoveal learning is much slower than foveal learning and that extrafoveal pattern representations are severely distorted. The latter distortions reveal the existence of limitations for the generalization of supervised learning over space.
Adult↗
Computational approaches to human pattern recognition.
This paper consolidates recent findings on how humans detect and recognize patterns and considers computational procedures which reflect observed performance. A multi-level correlation model for spatial information processing is proposed and used to interpret past results on human psychophysical performance.
Computer Graphics↗
A general correspondence approach to apparent motion.
A general framework is considered for how different features of image parts determine the perceived direction of apparent motion between these parts as a function of their internalized feature weights. It is shown how the compatibility and constraints between pairwise part correspondences also play important roles in the types of perceived motion between parts; this process is modelled via a multivariate constraint-satisfaction procedure.
Adult↗
Texture classification and segmentation algorithms in man and machines.
In this paper an attempt is made to review the types of processes developed by the author, and others, to classify and segment textures and to compare, where possible, such algorithms with what has been proposed to occur in human vision. In particular, this paper is concerned with evaluating the proposed constituent processes from a 'cognitive engineering' perspective where, in order to have an adequate model for either human or machine texture processing, complete segmentation or classification or both must be attained.
Algorithms↗
Perceptual spatial frequency--orientation surface: psychophysics and line element theory.
We have investigated the discriminability of gratings which simultaneously vary in spatial frequency and orientation. Thirteen and nine reference gratings were used with two observers, and bivariate discrimination probability surfaces were determined around each grating. These data were then fitted to a general bivariate Gaussian function. The results clearly demonstrate local separability in this log frequency and orientation discrimination domain. Our results also show that the factor contributing most to the non-Euclidean nature of such frequency/orientation discrimination is orientation anisotropia, although we also find some evidence for smaller changes in the associated Riemannian line-element at different frequency ranges. These results cast doubt upon claims for a "pseudo line-element" for frequency discrimination based upon the nonlinear outputs of a fixed set of detectors.
Discrimination, Psychological↗
On the discrimination of compound Gabor signals and textures.
We compared the discrimination of compound Gabor signals when presented pairwise and in isolation to that of textures consisting of an array of such signals. Textures composed of mirror-image compound Gabor signals were found indistinguishable even when the individual micropatterns were easily discriminated. No such blindness exists for textures composed of non-mirror-image compound Gabor signals. These results are analogous to earlier findings concerning the foveal and extrafoveal rapid discrimination of compound gratings [Rentschler and Treutwein, Nature 313, 308-310 (1985)]. This suggests that both texture perception and extrafoveal vision are, under the conditions of the given experiments, characterized by the absence of neural interactions that otherwise allow the analysis of form. Our results do also imply that the texton concept of Julesz can be extended to grey-level Gabor textures, irrespectively of whether the space domain (geometric features) or the spatial frequency domain (local spectral characteristics) representations of the micropatterns are referred to.
Discrimination, Psychological↗
Visual phase resolution for gray-scale textures.
Explore the source record for details and available documents.
Adult↗
Visual pattern recognition in humans. I. Evidence for adaptive filtering.
We have investigated how observers learn to classify compound Gabor signals as a function of their differentiating frequency components. Performance appears to be consistent with decision processes based upon the least squares minimum distance classifier (LSMDC) operating over a cartesian feature space consisting of the real (even) and imaginary (odd) components of the signals. The LSMDC model assumes observers form prototype signals, or adaptive filters, for each signal class in the learning phase, and classify as a function of their degree of match to each prototype. The underlying matching process can be modelled in terms of cross-correlation between prototype images and the input sample.
Form Perception↗
Is pattern masking predicted by the cross-correlation between signal and mask?
We sought to determine whether the cross-correlation between signals and masks consisting of Gaussian modulated grating patterns is consistent with observed psychophysical masking effects. Our experimental results support this relationship, and suggest further specifications for the determinants of pattern masking effects.
Form Perception↗
On the detection of signals in non-white noise.
We have studied the detection, by human observers, of suprathreshold bandlimited signals embedded at various locations in non-white, Gaussian filtered noise. Detection models based upon the direct cross-correlation between the signal and the noise image (matched filtering) cannot account for the results of our experiments. Our findings point instead at a detection process occurring at the level of signal decomposition, and jointly determined by: (a) the differential outputs of discrete, bandlimited spatial analyzers selectively responsive to different components of the signal; and (b) variable detection rules adaptively related to such outputs and to the type of signal information available to the observer.
Humans↗