An evaluation is made of the performance of two segmentation algorithms. The extreme difficulty of evaluating the quality of such segmentations, particularly on images of natural scenes, is discussed. The first algori...
详细信息
An evaluation is made of the performance of two segmentation algorithms. The extreme difficulty of evaluating the quality of such segmentations, particularly on images of natural scenes, is discussed. The first algorithm involves the symbolic labelling of pixels based upon cluster analysis of histograms and a probabilistic relaxation labelling algorithm. It is shown that serious segmentation problems are overcome when the algorithm is properly localized to overlapping subimages of data. Both algorithms are applied to fairly challenging machine-generated test images, as well as a natural image.
The k-syntactic similarity approach is couched in graphical representation terms and its ability to provide global recognition capability while retaining a low time complexity is explored. One potential application do...
详细信息
The k-syntactic similarity approach is couched in graphical representation terms and its ability to provide global recognition capability while retaining a low time complexity is explored. One potential application domain, that of composite shape decomposition into approximately convex subshapes, is described. This is shown to be equivalent to finding cycles within a particular graph. The approach yields valid decompositions in many cases where additional semantic considerations are necessary for proper analysis. The permissible graph structures representing composite shapes given a reasonable set of relations are determined. Experimental results on non-ideal data are given.
A new class of texture features based on the joint occurrences of gray levels at points defined relative to edge maxima are introduced. These features are compared with previous types of cooccurrence-based features, a...
详细信息
A new class of texture features based on the joint occurrences of gray levels at points defined relative to edge maxima are introduced. These features are compared with previous types of cooccurrence-based features, and experimental results are presented indicating that the new features should be useful for texture classification. In the second part, three simple methods of extracting texture primitives are compared. It appears that the simplest of these, thresholding at a fixed percentile, yields primitives that are quite effective in texture discrimination.
When radar and optical images are examined in detail, it is often found that the most distinguishable features of the two types of images are the shapes of the objects in the scenes. Therefore, edges can be used to ad...
详细信息
When radar and optical images are examined in detail, it is often found that the most distinguishable features of the two types of images are the shapes of the objects in the scenes. Therefore, edges can be used to advantages in the recognition and matching of objects. An edge extraction technique was developed and used to extract the salient outlines of objects of interest. This method also removes many of the edges extracted from the background and shadows around the objects.
The problem of learning in patternrecognition using imperfectly labeled patterns is considered. Using a probabilistic model for the mislabeling of the training patterns, the author discusses performance of the Bayes ...
详细信息
The problem of learning in patternrecognition using imperfectly labeled patterns is considered. Using a probabilistic model for the mislabeling of the training patterns, the author discusses performance of the Bayes and nearest neighbor classifiers with imperfect labels. Schemes are presented for training the classifier using both parametric and nonparametric techniques. Methods are developed for the correction of imperfect labels. To gain an understanding of the learning process, the author derives expressions for success probability as a function of training time for a one-dimensional increment error correction classifier with imperfect labels. Furthermore, feature selection with imperfectly labeled patterns is considered.
Quantitative analysis of lung tissue micrographs aims primarily at the determination of three-dimensional properties of the lung from two-dimensional micrographs. A computer-based approach to the determination of lung...
详细信息
Quantitative analysis of lung tissue micrographs aims primarily at the determination of three-dimensional properties of the lung from two-dimensional micrographs. A computer-based approach to the determination of lung tissue boundaries and triple points and end points from the micrograph is presented. The lung tissue boundaries are determined by histogram enhancement approach and zoom thresholding technique. The binary picture representing the lung tissue is then undergoing a thinning process to generate its skeleton from which the triple points and end points are automatically determined. The processing algorithms are developed into a software package implemented on a PDP-11/40 minicomputer.
In a previous paper, the authors present an O(kl multiplied by k2) depth-first tree search algorithm to find a montonic transformation such that the histogram of a transformed picture is closest to the desired histogr...
详细信息
In a previous paper, the authors present an O(kl multiplied by k2) depth-first tree search algorithm to find a montonic transformation such that the histogram of a transformed picture is closest to the desired histogram of a given picture in the L1 norm sense (least sum of absolute differences). The same problem using Ln norm is investigated in this paper. A breadth-first tree search algorithm with cut-off threshold is described, which has time complexity 0(k1 multiplied by k2).
The objective of computer vision system is to outline the objects in a picture and label them with an appropriate interpretation. The system consists of three different types of units, two of which are associative dat...
详细信息
The objective of computer vision system is to outline the objects in a picture and label them with an appropriate interpretation. The system consists of three different types of units, two of which are associative data memories implemented as relational databases. The short term memory (STM) contains the raw color picture data and the most current interpretations and deductions about the original scene. The long term memory (LTM) contains a detailed model of the scene under consideration. Results obtained with this system are presented.
A new distance measure is introduced for comparing polygonal curves which may be geometrically distorted relative to one another. The measure includes both directional and positional differences between segments and i...
详细信息
A new distance measure is introduced for comparing polygonal curves which may be geometrically distorted relative to one another. The measure includes both directional and positional differences between segments and is relatively insensitive to segmentation irregularity. Segment pairing is implemented with a conventional dynamic programming technique. Polygonal approximation allows great search reduction over pixel-based matching, thus making the technique practical for use on small computers. In addition, the measure is compatible with standard pruning techniques for even greater increases in speed. Some preliminary results in regard to handprinted character matching are shown.
Two means for numerically evaluating edge detectors, relative and absolute grading, were developed. Relative grading involves comparing an edge operator's output to the concensus decision of other operators. Absol...
详细信息
Two means for numerically evaluating edge detectors, relative and absolute grading, were developed. Relative grading involves comparing an edge operator's output to the concensus decision of other operators. Absolute grading compares any operator to a manually constructed key or target scene. A test of the ability of these techniques to eliminate operator and thresholding noise, improve threshold levels, system parameters and to guide operator improvement was carried out on six operators: Sobel, Robert's gradient, Roberts' square root gradient, gradient, range and Frei-Chen.
暂无评论