The image texture of objects possessing approximate radial symmetry is not effectively measured in a conventional rectangular coordinate frame. A method of texture measurement has been derived based on spatial depende...
详细信息
The image texture of objects possessing approximate radial symmetry is not effectively measured in a conventional rectangular coordinate frame. A method of texture measurement has been derived based on spatial dependence probability (SP) matrices formed from a moving polar reference frame. This method permits the resolution of texture into radial and tangential components. In radiocolloid liver images the effect of radial gradients makes texture measurements in a rectangular frame dependent on image size. This dependence was reduced in the tangential texture components of a trial set by using polar reference texture. 8 refs.
The problem examined is that of reconstructing a three-dimensional internal distribution using projection data taken with a number of suitably chosen cone-beam x-ray sources located outside the object of interest. For...
详细信息
The problem examined is that of reconstructing a three-dimensional internal distribution using projection data taken with a number of suitably chosen cone-beam x-ray sources located outside the object of interest. For a given exposure interval, the use of the cone-beam x-ray sources would provide information about the anatomical distribution within a volume rather than within a single slice. A times series of azimuthal scans taken with hardware consisting of electronically-switched stationary sources and detectors should provide sufficient data for fully four-dimensional image reconstructions that would permit detailed studies of the dynamic heart.
A rapid single-pass method is presented for processing a parallel-scan-encoded line drawing to obtain object size and shape information, and inter-object relationships in a picture. The procedure is efficient in stora...
详细信息
A rapid single-pass method is presented for processing a parallel-scan-encoded line drawing to obtain object size and shape information, and inter-object relationships in a picture. The procedure is efficient in storage requirements and needs the scan lines of data for boundary tracking and 2N words of core (for a scan line width of N) to extract shape features such as the ″necks″ of an object. The time and storage efficiency of shape feature extraction is achieved by using simple concepts from projective geometry. The technique is applicable to any picture containing objects (perhaps overlapping) with nontouching simply-connected boundaries.
The processing of images to extract regions, boundaries, and objects creates a spatial data base which contains large quantities of information consisting of objects, their attributes, their locations, and spatial rel...
详细信息
The processing of images to extract regions, boundaries, and objects creates a spatial data base which contains large quantities of information consisting of objects, their attributes, their locations, and spatial relationships. The problem of representing such spatial data in a uniform structure from which queries may be answered, commands may be carried out, and matching may be performed. The kinds of manipulations required in a spatial data base to answer queries about spatial data are discussed. It is then shown that matching of spatial data structures is a mathematical problem of finding homomorphisms from one spatial data structure to another.
A maximum likelihood formulation is given for the estimation of highly variable object boundaries in noisy images. This approach focuses on good stochastic models for data generation, and then optimum use of all the d...
详细信息
A maximum likelihood formulation is given for the estimation of highly variable object boundaries in noisy images. This approach focuses on good stochastic models for data generation, and then optimum use of all the data in the image. A framework is thus provided within which to view the good previously developed informal deterministic sequential approaches to boundary estimation and facilitates recognizing useful improvements which can be made. A dynamical systems model for boundary generation is introduced which is related to, but more useful than Fourier Series models. The treatment is completed by presenting an analytical approach for predicting the error to be expected in estimating highly variable boundaries.
The application of invariant hypothesis testing techniques to the classification (or detection) of spatially-distributed targets on the basis of radar observations is considered. Most powerful invariant solutions of t...
详细信息
The application of invariant hypothesis testing techniques to the classification (or detection) of spatially-distributed targets on the basis of radar observations is considered. Most powerful invariant solutions of the abstraction problem, when the multiple space cell observations are samples from a complex Gaussian process, are obtained for certain classes of problems. These include problems where the Doppler velocity spatial distributions are known, while the RCS spatial distributions are unknown. The Neyman-Pearson solution to problems of this type are not applicable since the hypotheses are composite;and uniformly most powerful solutions do not exist. However, by restricting the class of possible solutions to those which are invariant with respect to scale changes in received echo power, most powerful tests and the distribution of the most powerful invariant test statistics are found for a number of practical classification problems.
A method is described for scene representation from a region segmented image using shape descriptors and adjacency matrices. The shape descriptors are formed by piecewise linear approximations to the boundary of a reg...
详细信息
A method is described for scene representation from a region segmented image using shape descriptors and adjacency matrices. The shape descriptors are formed by piecewise linear approximations to the boundary of a region. Thus, each region is described by its shape characteristics. To illustrate that binary relationships must also be included in the scene representation, weighted adjacency matrices are formed for the scene. The combination of the shape characteristics and adjacency matrix form a concise representation of the scene but may also be used to reconstruct the major attributes of the scene. Several pictorial examples are given to illustrate the degree of scene representation possible using only shape descriptors and an adjacency matrix. Certain portions of the representation such as the adjacency matrix may be used as invariant features for scene matching. The total representation may also be used to reconstruct the salient features of the scene or as a basis for further analysis.
The quality of reconstructed ultrasonic images is limited by various processing and scanning steps in the reconstruction. Based on a simulation of ultrasonic ray paths through tissue-like models, certain limitations w...
详细信息
The quality of reconstructed ultrasonic images is limited by various processing and scanning steps in the reconstruction. Based on a simulation of ultrasonic ray paths through tissue-like models, certain limitations were evaluated, and reconstruction techniques were contrasted. Specifically, image fidelity was evaluated as a function of sample density;in the cases studied, the optimum number of projection was much less than pi /2 times the number of samples. Beamwidth deconvolution and spatial filtering significantly improved image fidelity in most cases. Reconstruction of attenuation, frequency dependent attenuation, and refractive index were compared for simple and structured targets. In most cases, attenuation reconstruction was not feasible except for small changes in the refractive index. Frequency dependent attenuation and refractive index reconstructions were comparable, with the later being slightly superior for simple targets.
computational techniques involving contrast enhancement and noise filtering on two-dimensional image arrays are developed based on their local mean and variance. These algorithms are nonrecursive and do not require th...
详细信息
computational techniques involving contrast enhancement and noise filtering on two-dimensional image arrays are developed based on their local mean and variance. These algorithms are nonrecursive and do not require the use of any kind of transform. They share the same characteristics in that each pixel is processed independently. Consequently, this approach has an obvious advantage when used in real-time digital imageprocessing applications and where parallel processor can be used. For both the additive and multiplicative noise, the a priori mean and variance of each pixel is derived from its local mean and variance. Then, the minimum mean square error estimator in its simplest form is applied to obtain the noise filtering algorithms. For multiplicative noise a statistical optimal linear approximation is made. Experimental results show that such an assumption yields a very effective filtering algorithm. Examples on images containing 256 multiplied by 256 pixels are given.
Digital processing of pulse echo ultrasound data has been utilized to study signal and image characteristics. Data was acquired by a dual processor, disk-based minicomputer system that permitted interactive processing...
详细信息
Digital processing of pulse echo ultrasound data has been utilized to study signal and image characteristics. Data was acquired by a dual processor, disk-based minicomputer system that permitted interactive processing and display. Fluctuating RF echoes obtained with a stationary beam from cardiac structures have been analyzed using Fourier techniques to show temporal changes during the cardiac cycle and to reveal variations in echo frequency content from disease. processing has also been carried out on matrices representing cross-sectional images that were produced by beam scanning. Small differences in signal amplitude have been emphasized by amplitude mapping based on histogram analysis of original images. Fourier domain filtering has been used to enhance spatial characteristics of ultrasonic images including tissue texture, reflector spacing, and structure boundaries. 6 refs.
暂无评论