Showing papers on "Kernel (image processing) published in 1984"

Matching Images Using Linear Features

Abstract: We describe techniques for matching two images or an image and a map. This operation is basic for machine vision and is needed for the tasks of object recognition, change detection, map up-dating, passive navigation, and other tasks. Our system uses line-based descriptions, and matching is accomplished by a relaxation operation which computes most similar geometrical structures. A more efficient variation, called the ``kernel'' method, is also described. We give results on complex aerial images which contain many image differences, caused by varying sun position, different seasons, and imaging environments, and also structural changes caused by man-made alterations such as new construction.

Image signal processor

Abstract: A high speed, multi-function and expandable image processing LSI (image signal processor) for realizing a gray level image processing technique is disclosed The architecture of the image signal processor can process a gray level image having 256 tones at a video rate (256×256 image, 6 MHz, non-interlace), allows expansion of a partial operation area (kernel) and can carry out various partial neighborhood operations The image signal processor is a partial parallel type image processing LSI which carries out a parallel operation by using the same number of processor elements as that of input pixel data used to produce one output pixel of data

Optical processing using outer-product concepts

Abstract: A row vector when left-multiplied by a column vector produces a two-dimensional rank-one matrix in an operation commonly called an outer product between the two vectors. The outer product operation can form the basis for a large variety of higher order algorithms in linear algebra, signal processing, and image processing. This operation can be best implemented in a processor having two-dimensional (2-D) parallelism and a global interaction among the elements of the input vectors. Since optics is endowed with exactly these features, an optical processor can perform the outer product operation in a natural fashion using orthogonally oriented one-dimensional (1-D) input devices such as acoustooptic cells. Algorithms that can be implemented optically using outer-product concepts include matrix multiplication, convolution/correlation, binary arithmetic operations for higher accuracy, matrix decompositions, and similarity transformations of images. Implementation is shown to be frequently tied to time-integrating detection techniques. These and other hardware issues in the implementation of some of these algorithms are discussed.

The Coupling Method for Solving Integral Equations

Abstract: This paper presents a new method to reduce integral operators of various classes to simpler operators, which often are just finite matrices. By this method the problem to find the inverse, generalized inverses, kernel and image of an integral operator is reduced for several cases to the corresponding problem for a finite matrix. The classes of integral operators dealt with include integral operators of the second kind with a finite rank or semi-separable kernel and also, which is more surprising, systems of Wiener-Hopf integral operators and singular integral operators with rational matrix symbols.

A technique of reinitialization for efficient simulation of large aquifers using the Discrete Kernel Approach

Abstract: The numerical technique of finite differences is commonly used to solve the partial differential equation describing saturated flow in aquifers (Boussinesq equation). This technique uses the known initial conditions at the beginning of the time step and the net excitations acting on the aquifer during the time step to compute the unknown heads at the end of the time step. This procedure requires the solution of a generally large system of simultaneous equations (equal to the number of nodes) at each simulation time step. Such an operation may become very costly for types of management problems where generally large aquifers have to be simulated numerous times at short time steps for a long time horizon. A technique which uses the discrete kernel approach and which requires the solution of the saturated flow equation for a small scanning grid system for a few time steps is presented. These solutions for few periods are used to simulate the heads in the total aquifer for long time horizons by a simple discrete convolution. The capability of the technique to handle point excitations and to predict point aquifer drawdowns are demonstrated. A cost comparison is made between the use of this technique and the conventional finite difference technique to solve a test aquifer simulation problem.

Speech recognition activation and deactivation method

Abstract: Speech recognition calculations are decreased by deactivating (or activating) a word in a grammar graph at the "kernel" level. A word is a sequence of acoustic kernels, each kernel a phoneme spectral vector with min-max duration data on a template.

One-Dimensional Systolic Arrays for Multidimensional Convolution and Resampling

Abstract: We present one-dimensional systolic arrays for performing two- or higher-dimensional convolution and resampling. These one-dimensional arrays are characterized by the fact that their I/0 bandwidth requirement is independent of the size of the convolution kernel. This contrasts with alternate two-dimensional array solutions, for which the I/0 bandwidth must increase as the kernel size increases. The proposed architecture is ideal for VLSI implementation—an arbitrarily large kernel can be handled by simply extending the linear systolic array with simple processors of the same type, so that one processor corresponds to each kernel element.

An algorithmic approach to non-homogeneous semi-markov processes

Abstract: De Dominicis 1979 suggested a generalization of the notion of semi-Markov process by introducing a suitable definition of the transition kernel of the process and proved a theorm on the existence and uniqueness of the solution of integral equation describing the process evolution. In this paper a quite general algorithm useful for the resolution of the evolution equation is performed; in addition an application to an actuarial problem is presented.

A kernel language for modules and abstract data types

Abstract: A small set of constructs can simulate a wide variety of apparently distinct features in modern programming languages. Using a kernel language called Pebble based on the typed lambda calculus with bindings, declarations, dependent types, and types as compile-time values, we show how to build modules, interfaces and implementations, abstract data types, generic types, recursive types, and unions. Pebble has a concise operational semantics given by inference rules.

Three-Dimensional And Four-Dimensional Convolutions By Coherent Optical Filtering

Abstract: Three-dimensional (3-D) and four-dimensional (4-D) signals can be replaced by two-dimensional (2-D) sequences of sectional images if sampling in one or two dimensions is practicable. The convolution of such sequences results in a sequence that under certain conditions is the sampled version of the result of 3-D or 4-D convolution. Thus higher dimensional convolution can be performed in two dimensions, e.g., by coherent-optical filtering. Because of the usually large space-bandwidth products (SBP) of such sequences, the parallel computing ability of optical filtering makes it very feasible. A relation between input signal size, SBP of the filter, and parameters of the coherent-optical setup allows one to estimate limits in pixel numbers for the input signal size, e.g., for comparison with digital techniques. Based on linear systems theory, the necessary filter for sequence convolution can be calculated from a given higher dimensional point response or transfer function. Experimental results for 3-D bandpass filtering and a 4-D correlation are discussed.© (1984) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Signal treatment apparatus based on aperture synthesis method

Abstract: PURPOSE:To obtain a good re-constituted image while suppressing side lobe, by taking convolution between the hologram signal obtained from a receiving wave and one obtained by multiplying kernel by window function having a wave form gentle in the slope thereof. CONSTITUTION:A scanning controller 14 controls a pulser 13 according to the indication from a computer CU21 and drives a probe to a living body 12 while cos hologram data and sin hologram data obtained by one emission of ultrasonic waves are respectively stored in memories 19, 20 and, thereafter, written in image memory 22. In the next step, the kernel data corresponding to the hologram data read from the memory 22 and converted to a complex number thereby while kernel is multiplied by window function having a wave form gentle in the skirt thereof and convolution integration with kernel multiplied by each hologram and window function is performed and the calculated result is accumulated in output buffer memory 23 to be displayed on TV monitor 26.

Homomorphic Processing in Ultrasonic NDE

Abstract: Modern signal processing techniques used in computing flaw impulse response are based on comparative analysis of a reference signal with the signal scattered from flaws. The underlying hypothesis in these techniques is that, in the noise free case, the scattered flaw signal is due to the linear convolution of the ultrasonic reference signal with the flaw impulse response. The flaw characterization problem thus reduces to determining the kernel of the convolution integral given the input and output time signals. The impulse response recovery (system identification) has been carried out both in the frequency and time domains for flaw characterization.1–3

A study of rectangular transforms for data compression

Abstract: A rectangular transform is defined and an algorithm for the same has been developed and implemented. The results show that a reasonable amount of data compression is possible. An additional amount of data eompression can also be obtained by modifying the kernel.

A two-dimensional adaptive image deblurring filter

Abstract: A recurrence relation is used to model image blurring, in place of the more conventional convolution. It is shown how this choice can lead to an efficient image restoration filter, by taking advantage of the recursive formulation.

Visual inspection using linear features

Abstract: This paper presents a method to inspect a printed circuit board. We use a feature-based system which matches symbolic descriptions of an object and a model. These symbolic descriptions are derived from line segments detected in an image. A model is built from a perfect board together with the expected defects. The matching technique is a variation of a line matching method ("Kernel" method) successfully applied at USC on aerial images. The results are then interpreted to generate a description of the match. Automated visual inspection Defects detection Industrial automation Graph matching Pattern recognition

Parallel processing of digital convolution using finite polynomial rings

Abstract: General method for the computation of digital convolution using residue arithmetic is developed. By the use of polynomial transforms over finite polynomial rings, convolution length and structural complexity typical for the FFT method are combined with simplest residue arithmetic. The whole structure is fully parallel and few building blocks composed of large VLSI ROM arrays are sufficient for high-speed pipelined computations of digital convolution.