The choice of whether to use autoregressive (AR) models to represent intensity or density data in imageprocessing.applications has often been made on heuristic grounds. The authors give a quantitative decision rule f...
详细信息
ISBN:
(纸本)0818606339
The choice of whether to use autoregressive (AR) models to represent intensity or density data in imageprocessing.applications has often been made on heuristic grounds. The authors give a quantitative decision rule for selecting between AR models for intensity versus density. They illustrate the usefulness of the decision rule by testing Seasat synthetic aperture radar imagery, conventional photographs of textures, and real-world scenes. Comparison is made between the decision rule results and results of goodness-of-fit tests.
The authors present a new approach to the stereo-matching problem that uses an edge-based description of images and a technique of prediction and verification of hypotheses. This approach was tested on a variety of in...
详细信息
ISBN:
(纸本)0818606339
The authors present a new approach to the stereo-matching problem that uses an edge-based description of images and a technique of prediction and verification of hypotheses. This approach was tested on a variety of indoor and outdoor scenes and also on scenes with industrial parts, and it appears to be fast, robust, and parallel-implementable.
A method is proposed for determining the shape of a rotating object in a laboratory environment. A camera and a turntable are the only equipment used. First, the optical flow generated on the camera image plane by tur...
详细信息
ISBN:
(纸本)0818606339
A method is proposed for determining the shape of a rotating object in a laboratory environment. A camera and a turntable are the only equipment used. First, the optical flow generated on the camera image plane by turning the object on the turntable and recording a set of images is calculated. Then, the calculated optical flow is used in simple relations to calculate parameters representing the shape of the object.
A description is given of a method for color image segmentation based on processing.the red, green, and blue components of the image simultaneously. The algorithm is based on a multichannel filtering model for the ima...
详细信息
ISBN:
(纸本)0818606339
A description is given of a method for color image segmentation based on processing.the red, green, and blue components of the image simultaneously. The algorithm is based on a multichannel filtering model for the image, where each channel represents an image in one of the three color components. This has the advantage that all of the correlations existing both within a given channel and between channels are automatically accounted for in the analysis. By combining the linear model with a 2-D Markov random field representing the occurrences of regions in the image, one is able to obtain an expression for the maximum a posteriori estimates of the regions. A brief description of the method, its application, and some possible simplifications of the algorithm are discussed.
A classification procedure is presented that differs from previous efforts in significant aspects. There is no sampling--the full image is used in the computation of class statistics. A null class is explicitly introd...
详细信息
ISBN:
(纸本)0818606339
A classification procedure is presented that differs from previous efforts in significant aspects. There is no sampling--the full image is used in the computation of class statistics. A null class is explicitly introduced to account for distributions not included in the training sets. The recursive partitioning proceeds top-down, limiting the number of partitions generated. Little attempt is made to 'optimize' the partitioning. A binary decision tree constructed from the partitioning is used to classify the full image.
A new solution to the problem of constructing a surface over a set of L cross-sectional contours is proposed. This surface is a polyhedron with triangular faces, obtained by pruning L-1 Delaunay triangulations between...
详细信息
ISBN:
(纸本)0818606339
A new solution to the problem of constructing a surface over a set of L cross-sectional contours is proposed. This surface is a polyhedron with triangular faces, obtained by pruning L-1 Delaunay triangulations between adjacent cross sections. This method can handle the case where there are several connected components in each cross section.
The conventional approach to the recovery of scene topography from a stereo pair of images is based on both the identification of distinctive scene features and the application of constraints imposed by the viewing ge...
详细信息
ISBN:
(纸本)0818606339
The conventional approach to the recovery of scene topography from a stereo pair of images is based on both the identification of distinctive scene features and the application of constraints imposed by the viewing geometry. A new prescription for recovering a relative-depth map is offered that involves integrating image irradiance profiles to find dense relative-depth profiles. The procedure neither matches image points (at least in the conventional sense) nor 'fills in' data to obtain the dense depth map. Although there are outstanding problems associated with depth discontinuities and image noise, the technique is effective.
作者：
Zhuang, XinhuaHaralick, Robert M.Univ of Michigan
Electrical Engineering & Computer Science Dep Ann Arbor MI USA Univ of Michigan Electrical Engineering & Computer Science Dep Ann Arbor MI USA
Given a set of corresponding points from a moving object in two perspective images, the authors give a complete theoretical and algorithmic solution to the two view motion problem. Experimentation on a computer verifi...
详细信息
ISBN:
(纸本)0818606339
Given a set of corresponding points from a moving object in two perspective images, the authors give a complete theoretical and algorithmic solution to the two view motion problem. Experimentation on a computer verifies the theory and the algorithm. The algorithm can be adjusted to the noise case;however, it needs an elaborate perturbation analysis.
A system is presented that takes a gray-level image as input, locates edges with subpixel accuracy, and links them into lines. Edges are detected by finding zero crossings in the convolution of the image with Laplacia...
详细信息
ISBN:
(纸本)0818606339
A system is presented that takes a gray-level image as input, locates edges with subpixel accuracy, and links them into lines. Edges are detected by finding zero crossings in the convolution of the image with Laplacian Gaussian masks. The implementation differs markedly from MIT's in that the masks are decomposed exactly into a sum of two separable filters instead of the usual approximation by a difference of two Gaussians (DOG). Subpixel accuracy is obtained through the use of the facet model. The zero-crossings obtained from the full resolution image using a space constant sigma for the Gaussian, and those obtained from the 1/n resolution image with 1/n pixel accuracy and a space constant of sigma /n for the Gaussian, are very similar, but the processing.times are very different. Finally, these edges are grouped into lines.
The problem of computing projections of digital images is considered. Algorithms that are suitable for implementation in general-purpose imageprocessing.and image analysis pipeline architectures are presented. The au...
详细信息
ISBN:
(纸本)0818606339
The problem of computing projections of digital images is considered. Algorithms that are suitable for implementation in general-purpose imageprocessing.and image analysis pipeline architectures are presented. The authors also propose some new pipeline configurations that achieve a remarkable degree of parallelism in the computation of projection data and, in fact, of many other geometrical descriptors of digital images. In particular, random access memories and other dedicated hardware devices are not needed in the algorithms. The effectiveness of the approach and feasibility of the proposed architectures are demonstrated by running the algorithms in commercially available short-pipelines for imageprocessing.and analysis. Examples are shown of the use of projection data for machine vision applications.
暂无评论