scispace - formally typeset
Search or ask a question

Showing papers on "Image gradient published in 1997"


Proceedings ArticleDOI
01 Feb 1997
TL;DR: It is shown that CCV’s can give superior results to color histogram-based methods for comparing images that incorporates spatial information, and to whom correspondence should be addressed tograms for image retrieval.
Abstract: Color histograms are used to compare images in many applications. Their advantages are efficiency, and insensitivity to small changes in camera viewpoint. However, color histograms lack spatial information, so images with very different appearances can have similar histograms. For example, a picture of fall foliage might contain a large number of scattered red pixels; this could have a similar color histogram to a picture with a single large red object. We describe a histogram-based method for comparing images that incorporates spatial information. We classify each pixel in a given color bucket as either coherent or incoherent, based on whether or not it is part of a large similarly-colored region. A color coherence vector (CCV) stores the number of coherent versus incoherent pixels with each color. By separating coherent pixels from incoherent pixels, CCV’s provide finer distinctions than color histograms. CCV’s can be computed at over 5 images per second on a standard workstation. A database with 15,000 images can be queried for the images with the most similar CCV’s in under 2 seconds. We show that CCV’s can give superior results to color his∗To whom correspondence should be addressed tograms for image retrieval.

931 citations


Proceedings ArticleDOI
17 Jun 1997
TL;DR: A novel boundary detection scheme based on "edge flow" that utilizes a predictive coding model to identify the direction of change in color and texture at each image location at a given scale, and constructs an edge flow vector.
Abstract: A novel boundary detection scheme based on "edge flow" is proposed in this paper. This scheme utilizes a predictive coding model to identify the direction of change in color and texture at each image location at a given scale, and constructs an edge flow vector. By iteratively propagating the edge flow, the boundaries can be detected at image locations which encounter two opposite directions of flow in the stable state. A user defined image scale is the only significant control parameter that is needed by the algorithm. The scheme facilitates integration of color and texture into a single framework for boundary detection.

258 citations


Patent
04 Aug 1997
TL;DR: A wide variety of modifications may be made to the source object in order to form modified stereo images as discussed by the authors, and these modifications may involve modifying one or more images relative to the original image.
Abstract: Methods, devices and systems are provided which produce a synthesized stereoscopic image of a source image by forming at least two images from a source image where at least one image has been modified relative to the source image such that the images have a different spacial appearance than each other. A wide variety of modifications may be made to the source object in order to form modified stereo images. These modifications may involve modifying one or more images relative to the source image. For example, at least one image may be magnified, reduced, or rotated in the X, Y and/or Z plane relative to the source image. Alternatively or in addition, the position of one or more elements of one of the images may also be changed relative to the source image. Alternatively or in addition, at least one of the images may be transformed relative to the source image using a function which alters the position of elements of the image along the X or Y axis relative to the source image. In one variation, the function is a distorting algorithm, for example an elliptical or aspheric distorting algorithm which is not dependent on the depth information of the source image.

184 citations


Journal ArticleDOI
TL;DR: A new method for combined color image segmentation and edge linking is proposed, which is modeled by a Gibbs random field and split and merged by a region-labeling procedure to enforce their consistency with the edge map.

121 citations


Patent
24 Nov 1997
TL;DR: In this article, a method and apparatus implementing dynamic thresholding techniques in processes that use image subtraction is described, where the subtracted image is separated into a positive difference image using a positive image map and a negative difference image by using a negative image map.
Abstract: A method and apparatus implementing dynamic thresholding techniques in processes that use image subtraction. The subtracted image is separated into a positive difference image using a positive image map and a negative difference image using a negative image map. A positive threshold map and a negative threshold map process a Gaussian filtered first image to generate respective positive and negative dynamic threshold images. The positive and negative threshold map mask out irrelevant areas and implement adaptive gain in pixel thresholds proportional to a difference or contrast between foreground or feature of interest and local background. Dynamic thresholding is a subtraction providing a subtracted positive image and a subtracted negative image. The subtracted positive and subtracted negative images are combined to provide a merged image which is binarized. Edge or Interface detection involves edge detection and true peak detection to find the edge peaks in a pre-image and post image. A binarization map converts edge peak pre and post images and the post image is then dilated to compensate any mis-registration. The dilated post image is subtracted from the binary edge pre-image to form an edge subtraction image. Pixels absent from the post image are assumed to be occluded by a feature of interest. The two binary images are merged, filtered, and connectivity is effected to obtain position and size of contiguous regions of interest.

101 citations


Journal ArticleDOI
TL;DR: This paper discusses an approach for detecting edges in color images that can easily accommodate concepts, such as multiscale edge detection, as well as the latest developments in vector order statistics for color image processing.
Abstract: This paper discusses an approach for detecting edges in color images. A color image is represented by a vector field, and the color image edges are detected as differences in the local vector statistics. These statistical differences can include local variations in color or spatial image properties. The proposed approach can easily accommodate concepts, such as multiscale edge detection, as well as the latest developments in vector order statistics for color image processing. A distinction between the proposed approach and previous approaches for color edge detection using vector order statistics is that, besides the edge magnitude, the local edge direction is also provided. Note that edge direction information is a relevant feature to a variety of image analysis tasks (e.g., texture analysis).

93 citations


Proceedings ArticleDOI
02 Nov 1997
TL;DR: A simple algorithm for tracking a person's head is presented, which tolerates full 360-degree rotation of the body as well as moderate amounts of occlusion, and it performs reacquisition of the subject.
Abstract: A simple algorithm for tracking a person's head is presented. A two-dimensional model, namely an ellipse, is used to approximate the head's contour. When a new image becomes available, a local search determines the position and size of the best ellipse by maximizing the normalized sum of the image gradient magnitude around the perimeter of the ellipse. The local search begins from a predicted position, using the head's velocity which eliminates the tracker's dependence upon maximum velocity. The tracker operates at 30 Hz and actively controls camera pan and tilt in order to track a person moving in a real environment. The algorithm tolerates full 360-degree rotation of the body as well as moderate amounts of occlusion, and it performs reacquisition of the subject.

92 citations


Patent
21 May 1997
TL;DR: In this article, a method and apparatus for segmenting image data into windows and for classifying the windows as typical image types includes making two passes through the image data, one for identifying windows and the other for labeling each of the pixels as a particular image type.
Abstract: A method and apparatus for segmenting image data into windows and for classifying the windows as typical image types includes making two passes through the image data. The method includes a step of making a first pass through the image data to identify windows and to record the beginning points and image types of each of the windows, and a step of making a second pass through the image data to label each of the pixels as a particular image type. The invention also includes a macro-detection method and apparatus for separating a scanline of image data into edges and image runs and for classifying each of the edges and image runs as standard image types. In the macro-detection method, image runs and edges are classified based on micro-detection results and based on image types of adjacent image runs.

85 citations


Journal ArticleDOI
TL;DR: Extracting textures from image sequences eliminates perspective distortions and removes highlights and reflections to produce high-quality multiresolution images with accurate color.
Abstract: Extracting textures from image sequences eliminates perspective distortions and removes highlights and reflections to produce high-quality multiresolution images with accurate color.

84 citations


Proceedings ArticleDOI
14 Oct 1997
TL;DR: A new quantization method for color images that uses a local error optimization strategy to generate near optimal quantization levels that are superior than those of other popular image quantization algorithms.
Abstract: This paper presents a new quantization method for color images. It uses a local error optimization strategy to generate near optimal quantization levels. The algorithm is simple to implement and produces results that are superior than those of other popular image quantization algorithms.

62 citations


Proceedings ArticleDOI
26 Oct 1997
TL;DR: A new algorithm, which uses a spatially variant measure of salience, which offers key improvements to a previous saturation feedback technique, and experimental results confirm that improved color image enhancement is achieved.
Abstract: One way of enhancing color image contrast is to feed back high-frequency spatial information from the saturation component into the luminance component. A new algorithm, which uses a spatially variant measure of salience, is presented. This method offers key improvements to a previous saturation feedback technique. Experimental results confirm that improved color image enhancement is achieved.

Patent
29 Oct 1997
TL;DR: In this paper, the amount of image data to be processed is reduced by dividing an image frame into a plurality of pixel blocks and expressing an image of each pixel block with a few of approximate color data and approximate color arrangement data.
Abstract: A data processing part in a printer controller performs multivalued dither processing whereby image data is processed for images of characters and thin lines with a high definition, and for pictures with a high color gradation, and the image quality deterioration at the boundary region between an edge region and a non-edge region can be avoided. The data processing part performs discrimination of an edge region, a non-edge region, and a boundary region between the edge and non-edge regions. Furthermore, the data processing part performs multivalued dither processing by using a prepared dot distribution type dither matrix for the edge region, a prepared dot concentration type dither matrix for the non-edge region and a prepared dot distribution and concentration mixing type dither matrix for the boundary region. Furthermore, the amount of image data to be processed is reduced by dividing an image frame into a plurality of pixel blocks and expressing an image of each pixel block with a few of approximate color data and approximate color arrangement data.

01 Jan 1997
TL;DR: This research provides an analysis that allows the prediction of the image-noisecharacteristics of systems for the capture of multispectral images and recommendations for detector characteristics and image processing for future systems are discussed.
Abstract: Peter D. BurnsSubmitted to the Center for ImagingScience in partial fulfillment of therequirements for Ph.D. degree atthe Rochester Institute of TechnologyMay 1997The design of a system for multispectral image capture will be influenced by the imagingapplication, such as image archiving, vision research, illuminant modification or improved(trichromatic) color reproduction. A key aspect of the system performance is the effect ofnoise, or error, when acquiring multiple color image records and processing of the data.This research provides an analysis that allows the prediction of the image-noisecharacteristics of systems for the capture of multispectral images. The effects of bothdetector noise and image processing quantization on the color information are considered,as is the correlation between the errors in the component signals.The above multivariate error-propagation analysis is then applied to an actual prototypesystem. Sources of image noise in both digital camera and image processing are related tocolorimetric errors. Recommendations for detector characteristics and image processing forfuture systems are then discussed.

Patent
01 Aug 1997
TL;DR: In this article, a linear transformation from a unit circle in the source image to the ellipse in the original image is calculated, and a pixel value in the destination image is determined by identifying pixels included in a support of the modified reconstruction filter.
Abstract: A source image may be manipulated by a computer to include special graphics effects. The source image may be used for still pictures or motion video applications and may include a single image or a sequence of images. The special graphics effects applied to the source image to create a desired destination image include various non-uniform scaling or arbitrary distortion functions. The effect has a corresponding affine transformation that approximates the non-uniform scaling function. This affine transformation is used to define an ellipse in the source image. A linear transformation from a unit circle in the source image to the ellipse in the source image is calculated. The linear transformation is used to modify a reconstruction filter in the source image. A pixel value in the destination image is determined by identifying pixels included in a support of the modified reconstruction filter in the source image in order to reduce artifacts in non-uniform scaling applications.

Patent
05 Dec 1997
TL;DR: In this article, a shape detection apparatus consisting of a video image producer, an image memory, an approximate position evaluator for evaluating the approximate position of the image of the target, and an edge point detector for defining plural sampling line windows having a predetermined length for detecting the image density at the detected approximate position, and detecting image density edge points in these sampling lines as the contour point positions of a target image.
Abstract: A shape detection apparatus able to accurately detect the image position of a target object even when the shape of the scanning window does not match the shape of the image of the target, and even when the scanning position of the scanning window does not precisely match the image position of the target is provided. This shape detection apparatus comprises a video image producer; an image memory; an approximate position evaluator for evaluating the approximate position of the image of the target; an edge point detector for defining plural sampling line windows having a predetermined length for detecting the image density at the detected approximate position of the image, and detecting image density edge points in these sampling line windows as the contour point positions of the target image; and a detection window setter for setting the contour lines of the target image from the detected contour point positions.

Proceedings ArticleDOI
26 Oct 1997
TL;DR: A color image edge detection algorithm is proposed based on the idea that use global color information to guide local gradient computation and edges are treated as transitions from one cluster to another.
Abstract: A color image edge detection algorithm is proposed based on the idea that use global color information to guide local gradient computation. The major chromatic components of an image are first extracted through cluster analysis. According to these color clusters, a set of linear chromatic transforms are generated. An appropriate chromatic transform is chosen for each pixel to maximize the gradient magnitude. In this way, edges are treated as transitions from one cluster to another. The algorithm is implemented and experimental results for real color images are included.

Book
13 Apr 1997
TL;DR: This paper presents a meta-analysis of statistical linear models used in image segmentation to derive Radial masks in line and edge detection and some approaches to image restoration.
Abstract: Preface 1. Introduction 2. Statistical linear models 3. Line detection 4. Edge detection 5. Object detection 6. Image segmentation 7. Radial masks in line and edge detection 8. Performance analysis 9. Some approaches to image restoration References Index.

Proceedings ArticleDOI
26 Oct 1997
TL;DR: This approach uses the morphological pyramid to form a scale space representation, offering a significant reduction in computational cost and avoids the over-segmentation problem of traditional fixed scale watershed algorithms.
Abstract: In this paper, we present a multiresolution implementation of the watershed segmentation algorithm. Our approach uses the morphological pyramid to form a scale space representation, offering a significant reduction in computational cost. In addition to increased efficiency, the multiresolution approach avoids the over-segmentation problem of traditional fixed scale watershed algorithms. As shown in the examples, the watershed pyramids produce edge maps corresponding to the desired scale without sacrificing accuracy in edge location.

Patent
18 Aug 1997
TL;DR: In this paper, a method of image filtering for carrying out picture element interpolation on a digital image signal obtained by sampling at a predetermined density an image signal representing an original image is presented.
Abstract: In a method of image filtering for carrying out picture element interpolation on a digital image signal obtained by sampling at a predetermined density an image signal representing an original image, the number of image signal components of the digital image signal is increased by interpolating, into the digital image signal, image signal components whose values are obtained on the basis of a predetermined operation using values of all the picture elements within the area covered with a filter and the coefficient of the filter corresponding to each of the picture elements. When an edge portion exists in the area covered with the filter, the predetermined operation is carried out after the image signal components for the picture elements representing the edge portion are corrected.

Patent
09 Apr 1997
TL;DR: In this paper, a region of substantially constant color within a first image is identified and a similar level of variation is applied to a second image, where the second image may be a video image or a computer generated image.
Abstract: Image data is processed by identifying a region of substantially constant color within a first image. Color variations, in terms of standard deviation about an average color, are analyzed and a similar level of variation is applied to a second image. The first image may be derived from cinematographic film, with color variations due to film grain. The second image may be a video image or a computer generated image and the application of a similar level of variation may reproduce noise similar to that present within the film due to grain. In this way, it is possible to add grain to non-filmed images so as to match said images to images derived from film.

Journal ArticleDOI
TL;DR: This letter deals with the color image edge enhancement issue using clustering ideas and based on the use of potential functions (Parzen windows) and employing the mountain clustering method and modifying it appropriately.
Abstract: This letter deals with the color image edge enhancement issue using clustering ideas and based on the use of potential functions (Parzen windows). Two algorithms are proposed. The first uses potential functions (PF's) and selects the output as the vector maximizing the PF. The second one elaborates further by employing the mountain clustering method and modifying it appropriately. Both algorithms are robust in the presence of noise, Gaussian and impulsive.

Patent
Masakazu Fujimoto1
09 Jan 1997
TL;DR: In this article, a document image processing device segments a set of pixels in a document into a plurality of regions and classifies the sizes of the characters in the segmented regions.
Abstract: A document image processing device segments a set of pixels in a document into a plurality of regions and classifies the sizes of the characters in the segmented regions. The characters are converted into a color image based on information indicating correspondence between size and color of the characters. An image is generated for an output document on the basis of the converted color image.

Patent
16 May 1997
TL;DR: In this paper, a wavelet transforming circuit is used to convert input image data into a plurality of band-limited image data, and an edge detecting circuit detects edge information in the image data based upon the band limited image data.
Abstract: An image processor accurately executes area separation using small block size image data. The image processor includes a wavelet transforming circuit which converts input image data into a plurality of band-limited image data, and an edge detecting circuit detects edge information in the image data based upon the band-limited image data. A character-area detecting circuit detects character areas in the image data from the distribution of the edge information, and an expansion/contraction circuit unifies the character areas based upon the detected character-area information.

Patent
17 Jul 1997
TL;DR: In this paper, an image processor includes spatial filters which execute parallel processing in units of component data of image data made up of a plurality of data, and an original image color determination unit selects if input image data from a color image input unit is to be processed as a monochrome or color image.
Abstract: An image processor includes spatial filters which execute parallel processing in units of component data of image data made up of a plurality of component data, and an original image color determination unit selects if input image data from a color image input unit is to be processed as a monochrome or color image. When it is selected that the input image data is to be processed as a monochrome image, the processing operations of the spatial filters for the image data are controlled to be executed stepwise.

Patent
14 Mar 1997
TL;DR: An image compression device comprises an image smoothing device which produces a smooth image by performing a smoothing process on an original image, an edge detection device that produces an edge image by subtracting said smoothed image from the original image.
Abstract: An image compression device comprises an image smoothing device which produces a smooth image by performing a smoothing process on an original image, an edge detection device which produces an edge image by subtracting said smoothed image from said original image, a fractal encoding device which encodes said smoothed image by fractal image compression, a reversible encoding device which encodes said edge image by reversible image compression, and an output device which combines the output of said fractal encoding means and the output of said reversible encoding means, and outputs said data as compressed data of the original image.

Journal ArticleDOI
TL;DR: A new multi-resolution method using color and motion information and shape model is developed to detect human faces in videophone QCIF sequences for efficient encoding using color segmentation and multiresolution propagation of a geometrical model.

Journal ArticleDOI
TL;DR: A novel color edge detection algorithm based on extraction of gradient discontinuities in the HSV color space is presented, capable of suppressing false edge detection in specular reflective regions.

Patent
14 Oct 1997
TL;DR: In this article, the rough image is generated based on non-compressed rough image data with a short reproduction waiting time and displayed on an image display device, where the corresponding full image data is processed for display so that a relevant full image replaces the displayed rough image, to be finally displayed.
Abstract: A digital still camera comprises a full image processor and a rough image processor. The full image processor generates compressed full image data as primary image data, while the rough image processor generates non-compressed rough image data as secondary image data whose data volume has been reduced by limiting pixels thereof. In a reproduction mode, rough image data is read from a memory and subjected to display processing. A rough image is generated based on non-compressed rough image data with a short reproduction waiting time and displayed on an image display device. Thereafter, corresponding full image data is processed for display so that a relevant full image replaces the displayed rough image data, to be finally displayed.

Journal ArticleDOI
TL;DR: Experimental results proved that the proposed selection criterion is effective in improving the performances of the detector over classical threshold methods.

Patent
Shin Mikuni1
12 Nov 1997
TL;DR: In this paper, a distribution rate of the respective color data of each of pixels in first image data and second image data is calculated in the case that the first image dataset is synthesized with the second dataset based on transmittance data of the first and second datasets.
Abstract: In an image processing apparatus, a plurality of images can be simply synthesized with each other without preparing a specific image such as an image extracting image. In this image processing apparatus, a distribution rate of the respective color data of each of pixels in first image data and second image data is calculated in the case that the first image data is synthesized with the second image data based on transmittance data of the first and second image data, and a value of color data of each pixel in a synthesized image is calculated, and then the first image data is synthesized with the second image data based upon this color data value.