Stochastic representation of discrete images by partial differential equations proposed earlier is used for filtering of noisy images. Comparisons among various representations and their relative effectiveness in mode...
详细信息
Stochastic representation of discrete images by partial differential equations proposed earlier is used for filtering of noisy images. Comparisons among various representations and their relative effectiveness in modeling the actual statistics of the images is shown by examples. Certain image decompositions resulting in fast filtering algorithms are given. The results are also compared with spatial averaging and Fourier domain Wiener filters.
A class of error-correcting tree automata is studied. Syntax errors on trees are defined in terms of five transformations - substitution, stretch, split, branch, and deletion. The distance between two trees is defined...
详细信息
A class of error-correcting tree automata is studied. Syntax errors on trees are defined in terms of five transformations - substitution, stretch, split, branch, and deletion. The distance between two trees is defined as the minimum cost sequence of error transformations needed to derive one from the other. This definition provides a similarity measure for syntactic patterns. For a given noisy input tree and a free language, the proposed generalized error-correcting tree automaton (GECTA) will produce a tree in the language using the minimum-distance criterion.
An application is presented of symmetric axis geometry to shape classification and description. Shapes are segmented into simplified segments, in which a sequential string of features are derived. These are based on b...
详细信息
An application is presented of symmetric axis geometry to shape classification and description. Shapes are segmented into simplified segments, in which a sequential string of features are derived. These are based on both width and axis properties, and two linear combinations of them, as well as their derivatives. A weighting measure is developed which evaluates the importance of these shape descriptors. Consequently, small perturbations are recognizable despite their generation of sizable symmetric axis arcs. A variety of other segment measures for shape are introduced.
Three transformations used in the analysis of tricolor natural scenes are analyzed: saturation, hue, and normalized color. All have nonremovable singularities, near which they are highly unstable. Given digital input,...
详细信息
Three transformations used in the analysis of tricolor natural scenes are analyzed: saturation, hue, and normalized color. All have nonremovable singularities, near which they are highly unstable. Given digital input, the distribution of their transformed values is highly nonuniform, characterized by spurious modes and gaps. These effects are quantified and illustrated. image segmentation techniques of edge detection, region growing, clustering, and region splitting are affected arbitrarily badly by such problems. Some stratagems are illustrated that help minimize the bad behavior.
A method is suggested for transforming a polygonal object into a binary relation whose clusters represent intuitively pleasing simple parts of the object. The graph theoretic clustering method determines near-cliques,...
详细信息
A method is suggested for transforming a polygonal object into a binary relation whose clusters represent intuitively pleasing simple parts of the object. The graph theoretic clustering method determines near-cliques, which are local regions of high compactness, and then forms clusters by merging those near-cliques having high enough overlap. Using this procedure on the shape of a colon as seen in an X-ray, it was found that for many colon examples, the parts determined by the clustering scheme corresponded well to the decomposition a human might make.
An image segmentation technique is proposed which uses a texture measure that counts the number of local extrema in a window centered at each picture point. Four gray level pictures are derived, each of which represen...
详细信息
An image segmentation technique is proposed which uses a texture measure that counts the number of local extrema in a window centered at each picture point. Four gray level pictures are derived, each of which represents a texture or gray level property of the original image. These intermediate pictures are used to derive starting points in each region to be segmented. The segmentation is completed by assigning each pixel to a starting point using a distance criteria. This process can be repeated in a hierarchical structure by using diminishing window sizes. Almost real-time video implementation seems feasible.
Given a particular encoding scheme, the complexity of a picture may be measured as the length of its encoding. An encoding scheme is described which is based on a hierarchy of resolution levels. The expected complexit...
详细信息
Given a particular encoding scheme, the complexity of a picture may be measured as the length of its encoding. An encoding scheme is described which is based on a hierarchy of resolution levels. The expected complexity of random pictures is computed. However, this method has some advantages. Here complexity tends to be more a function of shape than of size or position. Also, any prefix of the encoding string for a picture describes a simplified approximation to the picture. Potential applications are discussed.
The objective of this work is to create an algorithm that will generate a table of colors with maximum contrast among the colors. An algorithm for color generation is developed using as parameters the total number of ...
详细信息
The objective of this work is to create an algorithm that will generate a table of colors with maximum contrast among the colors. An algorithm for color generation is developed using as parameters the total number of colors to be produced, the number of colors to be grouped for maximum intra-group contrast, the number of saturation levels, and an arbitrary rotation of the selected colors preserving all the other relationships. The Hering opponent color theory restricted to the red-green and blue-yellow dimensions yields a two-dimensional color model which is used as the basis of the color selection process.
This paper concerns methods for indexing areas in two-dimensional array data. A method for naming subpictures from raster scan image data is presented with notation that eases their subsequent storage access. Equation...
详细信息
This paper concerns methods for indexing areas in two-dimensional array data. A method for naming subpictures from raster scan image data is presented with notation that eases their subsequent storage access. Equations are given for converting each subpicture name into a storage-location pointer. A function ″NUMERIC″ to take an area name and return its unique relative offset on auxiliary storage is presented. Algorithms enabling efficient retrieval of subpicture areas from sequential and direct access files are presented.
Two sets of features - perceptual features and co-occurrence features - were used in classification experiments. The ability of each set was verified in a pilot experiment for classification of 16 typical natural text...
详细信息
Two sets of features - perceptual features and co-occurrence features - were used in classification experiments. The ability of each set was verified in a pilot experiment for classification of 16 typical natural texture patterns. In the main study several practical problems concerning usage of textural features are considered by using pairs of multispectral images sensed at different altitudes on the same day. It was verified that adequate features could be selected by visual estimation. Finally, the effects of spectral bands and resolution are mentioned. This work was applied to agricultural land use classification.
暂无评论