Showing papers by "David P. Casasent published in 1992"
TL;DR: This new minimum noise and correlation energy filter approach introduces the concept of using the spectral envelope of the training images and the noise power spectrum to obtain a tight bound to the energy minimization problem that is associated with distortion-invariant filters in noise.
Abstract: A new distortion-invariant optical correlation filter to produce easily detectable correlation peaks in the presence of noise and clutter and to provide better intraclass recognition is presented. The basic ideas of the minimum variance synthetic discriminant function correlation filter (which minimizes noise variance in the output correlation peak/plane) and the minimum average correlation energy filter (which minimizes the average correlation plane energy over all the training images) are unified in a new filter that produces sharp correlation peaks while maintaining an acceptable signal-to-noise ratio in the correlation plane output. This new minimum noise and correlation energy filter approach introduces the concept of using the spectral envelope of the training images and the noise power spectrum to obtain a tight bound to the energy minimization problem that is associated with distortion-invariant filters in noise while allowing the user a variable parameter to adjust depending on the noise or clutter that is expected. We present the mathematical basis for the minimum noise and correlation energy filter and the initial simulation results.
149 citations
TL;DR: The Gabor coefficients are found to be excellent for object detection and robust to object distortions and contrast differences.
Abstract: We consider wavelet and Gabor transforms for detection of candidate regions of interest in a 2-D scene. We generate wavelet and Gabor coefficients for each spatial region of a scene using new linear combination optical filters to reduce the output dimensionality and to simplify postprocessing. We use two sets of wavelet coefficients as indicators of edge activity to suppress background clutter. The Gabor coefficients are found to be excellent for object detection and robust to object distortions and contrast differences. We provide insight into the selection of the Gabor parameters.
93 citations
TL;DR: An optical correlator implementation of the morphological hit-miss transform that provides improved recognition of objects in clutter compared with standard optical pattern-recognition correlator techniques is presented.
Abstract: We present an optical correlator implementation of the morphological hit-miss transform. This provides improved recognition of objects in clutter compared with standard optical pattern-recognition correlator techniques. The hit-miss transform is modified to use rank-order filtering since this gives better performance in noise and clutter. By varying the correlation plane threshold, we can perform dilations, rank-order filters, and erosions on the same multifunctional coherent optical correlator system. We quantify the thresholds required for generic object part recognition and provide simulated and optical laboratory data.
68 citations
TL;DR: The original minimum average correlation energy (MACE) filter is addressed by using a new database and including noise performance, depression angle, and resolution effects on the number of training set images that are required.
Abstract: The original minimum average correlation energy (MACE) filter is addressed by using a new database (strategic relocatable objects, missile launchers) and including noise performance, depression angle, and resolution effects on the number of training set images that are required. Major attention is given to our new MACE filter algorithms for distortion-invariant pattern recognition: shifted-MACE filters (to suppress large false correlation peaks), minimum variance-MACE filters (for improved noise performance), multiple symbolic encoded filters (to reduce the effect of false correlation peaks), and Gaussian-MACE filters (to improve noise performance and intraclass recognition and reduce the training set size).
49 citations
TL;DR: It is shown that one must employ new recollection vector encoding techniques to improve storage density, else the standard direct storage nearest neighbor processor is preferable.
41 citations
TL;DR: The three optical information processing techniques of detection, recognition, and identification can and should be combined to achieve the best benefits of each.
Abstract: The three optical information processing techniques of detection, recognition, and identification can and should be combined to achieve the best benefits of each. All methods are required for difficult pattern recognition problems. We consider the identification of multiple objects in the field of view in clutter. A morphological correlator is used to achieve detection. Hierarchical and symbolic pattern recognition correlators can also achieve detection as well as recognition. For very large class probems, feature extractors are required for identification, but first require detection. For difficult multiclass discrimination problems, neural net methods (rather than linear discriminant functions) are needed for identification.
30 citations
01 Nov 1992
TL;DR: A new gray-scale clutter reduction morphological algorithm for low clutter cases and a new algorithm for high clutter cases are presented and binary structuring elements are found to be adequate; this is very attractive for ourgray-scale morphology decomposition algorithm and its optical implementation.
Abstract: We consider morphological processing for clutter reduction and object detection. For detection, we compare a binary and gray-scale Hit-Miss Transform and find that the binary operator is preferable. For clutter reduction, we find gray-scale morphology to be preferable. We present a new gray-scale clutter reduction morphological algorithm for low clutter cases and a new algorithm for high clutter cases. In all morphological processing, we find binary structuring elements to be adequate; this is very attractive for our gray-scale morphology decomposition algorithm and its optical implementation.© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.
22 citations
01 Apr 1992
TL;DR: How gray scale morphology can be implemented on an optical correlator system using a threshold decomposition algorithm is described and ways to reduce the number of intermediate processing steps required are discussed.
Abstract: There is much work concerning morphological image processing, both binary and gray scale. Almost all implementations to date are performed electronically on standard computers, specialized processors, or specialized hardware. Prior work has described implementation of binary morphology on an optical processor, as well as indicating the relative merits of using an optical system. However, the restriction to binary morphology on an optical system has required that gray scale problems be reduced to binary morphology solutions using judiciously chosen binarization thresholds. This paper describes how gray scale morphology can be implemented on an optical correlator system using a threshold decomposition algorithm. A series of thresholded binary correlations are formed optically and summed on a CCD detector array or spatial light modulator, to produce the output morphologically processed gray scale image. The speed this optical system is much faster than 30 gray scale images per second. The details of the architecture used to implement threshold decomposition on an optical system is described, and issues relating to the implementation of binary morphology on an optical system are discussed. The threshold decomposition algorithm is discussed with attention to ways to reduce the number of intermediate processing steps required.
16 citations
TL;DR: A new neural net is described that can easily and cost-effectively accommodate multiple objects in the field of view in parallel and the use of a correlator achieves shift invariance and accommodates multipleObjects in parallel.
Abstract: A new neural net is described that can easily and cost-effectively accommodate multiple objects in the field of view in parallel. The use of a correlator achieves shift invariance and accommodates multiple objects in parallel. Distortion-invariant filters provide aspect-invariant distortion. Symbolic encoding, the use of generic object parts, and a production system neural net allow large class problems to be addressed. Optical laboratory data on the production system inputs are provided and emphasized. Test data assume binary inputs, although analog (probability) input neurons are possible.
16 citations
01 Nov 1992
TL;DR: A hierarchical inference set of such filters allows scene analysis on a unified multifunctional correlator architecture that has applications in robotics, computer vision, optical character recognition, reconnaissance, and target recognition.
Abstract: We consider new correlation filters for all levels of scene analysis such as clutter reduction and object detection, recognition, and identification. A hierarchical inference set of such filters allows scene analysis on a unified multifunctional correlator architecture. It has applications in robotics, computer vision, optical character recognition, reconnaissance, and target recognition. Present digital processors can now achieve correlations in real time and hence such filters are of importance.© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.
12 citations
TL;DR: This paper is intended to study the suitability of using optical processing techniques for the signal Gabor and wavelet analysis of one- and two-dimensional signals and images.
Abstract: Recent development in vision and image understanding related study reveals that a signal decomposition before processing may provide enormous useful information about the signal. Various signal decomposition models, such as the Gabor and wavelet transforms have been proposed. While the Gabor signal expansion creates a fixed resolution space-frequency signal representation, the wavelet transform provides a multi-resolution signal space-scale decomposition. Digital implementation of these transforms are computationally intensive both because of the nature of the coordinate-doubling of the transforms and due to the large quantity of convolution/correlation operations to be performed. Optics with its inherent parallel processing capability has been applied to many useful linear signal and image transformations for feature analysis and extraction. This paper is intended to study the suitability of using optical processing techniques for the signal Gabor and wavelet analysis. Gabor and wavelet transforms of both one- and two-dimensional signals and images are discussed. System parameters and limitation are analyzed. Preliminary experimental results are presented.
TL;DR: A new symbolic correlator and production system optical neural net to handle multiple objects in the field of view and the implementation of a wide variety of neural net functions on a multifunctional hybrid optical/digital neural net system are presented.
TL;DR: A neural network solution to the data association problem in multitarget tracking that uses position and velocity measurements of the targets over two consecutive time frames and shows that the network performs well when track data are corrupted by significant noise.
Abstract: A neural network solution to the data association problem in multitarget tracking is presented. It uses position and velocity measurements of the targets over two consecutive time frames. A quadratic neural energy function, which is suitable for an optical processing implementation, results. Simulation results using realistic target trajectories with target measurement noise including platform movement or jitter are presented. The results show that the network performs well when track data are corrupted by significant noise. Several possible optical neural network architectures to implement this algorithm are discussed, including a new all-optical matrix–vector multiplication approach. The matrix structure is employed to allow binary–ternary spatial light modulators to be used.
01 Mar 1992
TL;DR: An optical realization that uses complex-valued weights, optical intensity detectors, and an additional input neuron to achieve piecewise conic decision surfaces (rather than the piecewise linear surfaces that the ACNN produces) is discussed.
Abstract: A neural network pattern classifier is presented. Its decision boundaries are formed from segments of conic sections which allows it to achieve improved performance over piecewise linear neural network classifiers, such as our earlier adaptive clustering neural network (ACNN). We discuss an optical realization that uses complex-valued weights, optical intensity detectors, and an additional input neuron to achieve piecewise conic decision surfaces (rather than the piecewise linear surfaces that the ACNN produces).
01 Nov 1992
TL;DR: A unified correlator architecture is used with inference filters that hierarchically process the input image scene to perform detection, enhancement, recognition, and finally identification.
Abstract: We consider the problem of identifying each of multiple objects in a scene with object distortions and background clutter present. A unified correlator architecture is used with inference filters that hierarchically process the input image scene to perform detection, enhancement, recognition, and finally identification. The different levels of the processor use various processing techniques: hit-miss rank-order and erosion/dilation morphological filtering, distortion-invariant filtering, feature extraction, and neural net classification.
01 Jan 1992
TL;DR: This chapter presents a hybrid optical/digital neural nets (NN) architecture that combines the advantageous properties of both optical and digital technology.
Abstract: Publisher Summary This chapter presents a hybrid optical/digital neural nets (NN) architecture that combines the advantageous properties of both optical and digital technology. No NN architecture can be separated from its algorithm and application. There are five different types of NNs, namely, associative processors, optimization NNs, symbolic correlator, production system NNs, and adaptive learning NNs. Many of these represent new NN algorithms. Many NNs use iconic neurons. For 500 × 500 pixel images, this requires an enormous number of neurons. A preferable approach is to use feature space neurons where each input neuron is associated with a specific image feature and the neuron's excitation value is proportional to that feature's value. Many feature spaces can be used such as Fourier coefficients, wedge ring sampled Fourier features, and moments. The three major advantages of feature space neurons are a significant reduction in the number of input neurons, a neuron space with some distortion invariance, and a significant reduction in training.
TL;DR: Use of a hierarchy of correlation filters allows one to address all levels of model based vision including detection, segmentation, and classification, which significantly reduces the object geometry and contrast search space required.
TL;DR: This work considers the classification of multiple objects in a scene with distortion and clutter present and proposes a hierarchical/inference approach using correlation NNs for low-level operations and new classifier NNs with higher-order decision surfaces for the final decision NNs.
Abstract: We consider the classification of multiple objects in a scene with distortion and clutter present. Our opinions on the role for neural nets (NNs) in this application and the different properties that NNs must have to address this problem are advanced. A hierarchical/inference approach is suggested using correlation NNs for low-level operations and new classifier NNs with higher-order decision surfaces for the final decision NNs. Our concern is NN capacity and performance (in noise). Our capacity guidelines advanced concern the number of neurons, use of analog neurons, Ho-Kashyap (HK) NNs, and two new NNs with higher-order decision surfaces. Our noise performance guidelines advanced concern the number of neuron layers, hidden-layer neuron encoding, and robust HK NNs.
TL;DR: This work details new generalized rotation-invariant filters with initial test results, and briefly reviews the hierarchical inference approach to scene analysis, and details the use of morphological hit-miss and MINACE filters for detection.
Abstract: Our research group at Carnegie Mellon University has devised a large number of new distortion-invariant optical correlation filters. We briefly review our hierarchical inference approach to scene analysis, our minimum noise and correlation energy (MINACE) filters, the use of morphological hit-miss (rank-order) and MINACE filters for detection, and the use of MINACE filters with different control parameters c for all three levels of scene analysis (detection, recognition, identification). We then detail new generalized rotation-invariant filters with initial test results.© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.
TL;DR: Two new computer generated hologram (CGH) elements for optical processing are reported on, including a one-to-many optical interconnection element and an element to provide separate 1-D collimation of the laser diodes in an array.
Abstract: We report on two new computer generated hologram (CGH) elements for optical processing. They are a one-to-many optical interconnection element (that allows analog weights, high efficiency, and is not restricted to a regularly spaced grid) and an element to provide separate 1-D collimation of the laser diodes in an array (with high efficiency). Error diffusion encoding and multilevel phase CGHs are used to achieve high accuracy and high efficiency. Simulations are used to show the advantage of error diffusion (ED) encoding. Optical laboratory data are included to show the feasibility of the elements and the validity of our simulator.
07 Jun 1992
TL;DR: Work at Carnegie Mellon University is emphasized and includes new hyperspherical Ho-Kashyap neural nets and new piecewise quadratic neural nets.
Abstract: Several recent advances are described that use neural-network methods to produce the higher-order decision surface required for difficult pattern recognition discrimination problems. Work at Carnegie Mellon University is emphasized and includes new hyperspherical Ho-Kashyap neural nets and new piecewise quadratic neural nets. Also addressed are Fourier neural-net interconnections to handle multiple objects and achieve morphological, image processing, and enhancement functions. >
01 Mar 1992
TL;DR: The adaptive clustering neural net and robust Ho-Kashyap (HK-2) associative processor (AP) are the neural networks considered in detail.
Abstract: We consider analog neural network implementations (using VLSI or optical technologies) with limited accuracy and various noise and nonlinearity error sources. Algorithms and techniques to achieve high performance (good recognition P' c % and large storage capacity) on such systems are considered. The adaptive clustering neural net (ACNN) and robust Ho-Kashyap (HK-2) associative processor (AP) are the neural networks considered in detail.
01 Mar 1992
TL;DR: A new morphological skeletonization algorithm, a Hough Transform together with morphological operations, morphological erosions with directional structuring elements, and new parameters to describe and distinguish textures are developed.
Abstract: Morphological processing combined with other techniques is used to analyze disordered structures. Disordered structures can consists of a number of objects of a given shape (the task is to determine the number of objects and their length and orientation distribution) or texture (this occurs when the number of particles is large) and the task is to describe the texture and to discriminate different textures. To solve such problems, we employ a new morphological skeletonization algorithm, a Hough Transform together with morphological operations, morphological erosions with directional structuring elements, and develop new parameters to describe and distinguish textures. Our algorithms can be implemented in digital or optical processors.© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.
Book•
01 Jan 1992
16 Nov 1992
TL;DR: Optical processors can perform the required operations for the various levels of a hierarchical/ inference computer vision system for scene analysis on a multifunctional programmable optical architecture.
Abstract: Optical processors can perform the required operations for the various levels of a hierarchical/ inference computer vision system for scene analysis (detection, enhancement, recognition, feature extraction, and classification) on a multifunctional programmable optical architecture.
07 Jun 1992
TL;DR: The author shows that the neural nets considered allow some such effects to be included inherently in the neural net synthesis algorithm and that the effect of the other error sources can be trained out by proper selection of neural net design parameters.
Abstract: Various errors, including analog accuracy, nonlinearities, and noise, are present in all neural networks. The author considers their effects in training and testing on two different pattern recognition neural nets. He shows that the neural nets considered allow some such effects to be included inherently in the neural net synthesis algorithm and that the effect of the other error sources can be trained out by proper selection of neural net design parameters. Multiclass distortion-invariant pattern recognition neural nets are considered. The results are applicable to analog VLSI and optical neural nets. >
16 Dec 1992
TL;DR: Many seemingly diverse optical processing operations are possible on one new multifunctional optical architecture and are described and provided examples of each for scene analysis.
Abstract: Many seemingly diverse optical processing operations are possible on one new multifunctional optical architecture. We describe these operations and provide examples of each for scene analysis. L© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.
01 Feb 1992
TL;DR: An optical correlator with banks of distortion invariant hierarchical/inference filters appears to be an ideal adjunct to other OCR techniques (AI, parsing, context, use of lexicons, etc.).
Abstract: An OCR machine printed problem is selected as an example of a large class pattern recognition problem. We consider discrimination of alpha and numeric fields, recognition of all numbers, recognition of key words (street suffixes, personal titles), state/city/street names, etc. These operations are performed on destination address blocks (DABs) in the face of numerous variations in the type face (laser writer, dot matrix, typewriter, etc.), font, data drop out (due to printing errors), point size, +/- 5 degree(s) rotations, etc. An optical correlator with banks of distortion invariant hierarchical/inference filters appears to be an ideal adjunct to other OCR techniques (AI, parsing, context, use of lexicons, etc.).© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.