Showing papers on "Pixel published in 1991"

Watersheds in digital spaces: an efficient algorithm based on immersion simulations

Abstract: A fast and flexible algorithm for computing watersheds in digital gray-scale images is introduced. A review of watersheds and related motion is first presented, and the major methods to determine watersheds are discussed. The algorithm is based on an immersion process analogy, in which the flooding of the water in the picture is efficiently simulated using of queue of pixel. It is described in detail provided in a pseudo C language. The accuracy of this algorithm is proven to be superior to that of the existing implementations, and it is shown that its adaptation to any kind of digital grid and its generalization to n-dimensional images (and even to graphs) are straightforward. The algorithm is reported to be faster than any other watershed algorithm. Applications of this algorithm with regard to picture segmentation are presented for magnetic resonance (MR) imagery and for digital elevation models. An example of 3-D watershed is also provided. >

Multiple resolution segmentation of textured images

Abstract: A multiple resolution algorithm is presented for segmenting images into regions with differing statistical behavior. In addition, an algorithm is developed for determining the number of statistically distinct regions in an image and estimating the parameters of those regions. Both algorithms use a causal Gaussian autoregressive model to describe the mean, variance, and spatial correlation of the image textures. Together, the algorithms can be used to perform unsupervised texture segmentation. The multiple resolution segmentation algorithm first segments images at coarse resolution and then progresses to finer resolutions until individual pixels are classified. This method results in accurate segmentations and requires significantly less computation than some previously known methods. The field containing the classification of each pixel in the image is modeled as a Markov random field. Segmentation at each resolution is then performed by maximizing the a posteriori probability of this field subject to the resolution constraint. At each resolution, the a posteriori probability is maximized by a deterministic greedy algorithm which iteratively chooses the classification of individual pixels or pixel blocks. The unsupervised parameter estimation algorithm determines both the number of textures and their parameters by minimizing a global criterion based on the AIC information criterion. Clusters corresponding to the individual textures are formed by alternately estimating the cluster parameters and repartitioning the data into those clusters. Concurrently, the number of distinct textures is estimated by combining clusters until a minimum of the criterion is reached. >

Physical performance characteristics of spiral CT scanning.

Abstract: CT scanning in spiral geometry is achieved by continuously transporting the patient through the gantry in synchrony with continuous data acquisition over a multitude of 360-deg scans. Data for reconstruction of images in planar geometry are estimated from the spiral data by interpolation. The influence of spiral scanning on image quality is investigated. Most of the standard physical performance parameters, e.g., spatial resolution, image uniformity, and contrast, are not affected; results differ for pixel noise and slice sensitivity profiles. For linear interpolation, pixel noise is expected to be reduced by a factor of 0.82; reduction factors of 0.81 to 0.83 were measured. Slice sensitivity profiles are changed as a function of table feed d, measured in millimeters per 360-deg scan; they are smoothed as the original profile is convolved with the object motion function. The motion function is derived for linear interpolation that constitutes a triangle with a base line width of 2d and a maximal height equal to 1/d. Calculations of both the full width at half-maximum and the shape of the profiles were in good agreement with experimental results. The effect of the widened profiles, in particular of their extended tail ends, on image quality is demonstrated in phantom measurements.

Method and apparatus for selectively encoding and decoding a digital motion video signal at multiple resolution levels

Abstract: At least one selected image of a digital video signal is encoded at multiple levels of resolution. For each level of resolution, a correction image is formed by subtracting the value of each pixel in a reference image of that resolution level from the value of each corresponding pixel in the selected image of that resolution level. The correction image is quantized and encoded. A decoder decodes the encoded quantized correction image for each resolution level. The value of each pixel in the decoded correction image having the lowest resolution is added to the value of each corresponding pixel of a reference image having the same resolution to form a result image of the lowest resolution. This result image is expanded to the next higher level of resolution. The value of each pixel in the expanded result image is added to the value of each corresponding pixel in the decoded correction image of the same resolution level to form a result image of that resolution level. This process is continued until each pixel value in the expanded result image, which has been expanded to full resolution, is added to the value of each corresponding pixel in the full resolution decoded correction image to form the final image which is stored in memory for subsequence display.

Detecting small, moving objects in image sequences using sequential hypothesis testing

Abstract: An algorithm is proposed for the solution of the class of multidimensional detection problems concerning the detection of small, barely discernible, moving objects of unknown position and velocity in a sequence of digital images. A large number of candidate trajectories, organized into a tree structure, are hypothesized at each pixel in the sequence and tested sequentially for a shift in mean intensity. The practicality of the algorithm is facilitated by the use of multistage hypothesis testing (MHT) for simultaneous inference, as well as the existence of exact, closed-form expressions for MHT test performance in Gaussian white noise (GWN). These expressions predict the algorithm's computation and memory requirements, where it is shown theoretically that several orders of magnitude of processing are saved over a brute-force approach based on fixed sample-size tests. The algorithm is applied to real data by using a robust preprocessing procedure to eliminate background structure and transform the image sequence into a residual representation, modeled as GWN. Results are verified experimentally on a variety of video image sequences. >

Rough surface profiler and method

Abstract: A method of profiling a rough surface of an object includes moving the object along a z axis so that a highest point of the rough surface is optically aligned with and outside of the focus range of a solid-state imaging array. An interferogram of the rough surface then is produced by means of a two beam interferometer. The solid-state imaging array is operated to scan the rough surface along x and y axes to produce intensity data for each pixel of the solid-state imaging array for a plurality of frames each shifted from the other by a preselected phase difference. The modulation for each pixel is computed from the intensity data. The most recently computed modulation of each pixel is compared with a stored prior value of modulation of that pixel. The prior value is replaced with the most recently computed value if the most recently computed value is greater. The object is incrementally moved a selected distance along the z axis, and the foregoing procedure is repeated until maximum values of modulation and corresponding relative height of the rough surface are obtained and stored for each pixel.

Digital halftoning with space filling curves

Abstract: This paper introduces a new digital halftoning technique that uses space filling curves to generate aperiodic patterns of clustered dots. This method allows the parameterization of the size of pixel clusters, which can vary in one pixel steps. The algorithm unifies, in this way, the dispersed and clustered-dot dithering techniques.

High efficiency light valve projection system with decreased perception of spaces between pixels and/or hines

Abstract: A light valve such as an active matrix LCD between crossed polarizers, utilizing, for instance, individual transistors to control each "pixel area" of the LCD and storage elements to store video signal data for each pixel, with optically shielded "dead spaces" between pixels to eliminate electric field crosstalk and non-information-bearing light bleed through, is illuminated with a bright independent light source which creates a video image projected via specialized projection optics onto an internal or external screen without distortions, regardless of the angle of projection onto the screen. Use of heat sinks, IR reflective coatings, heat absorbing optics, optional fluid and a thermistor controlled pixel transistor bias voltage injection servo circuit stabilizes image performance, maintaining accurate color and contrast levels as the LCD changes temperature. In one embodiment of the invention, use of a multi-color LCD with a stepped cavity, producing different thicknesses of LCD for the different wavelengths that pass through it, allows a linear correspondence between the wavelengths passing through the LCD to produce true black, high contrast and CRT-like color rendition. A dichroic mirror arrangement is used to overlap differently colored pixels in the projected image. Use of lens arrays to expand pixels, where necessary, eliminating spaces between pixels, creating a continuous image with no apparent stripes or dots. A special venetian-blind type of screen is also disclosed and methods for using the system to view three-dimensional video are also explained.

Adaptive motion compensation for digital television

Abstract: A method and apparatus are provided for processing digital video signals for transmission in a compressed form. A set of pixel data is compressed without motion compensation to provide a first compressed video signal. The pixel data is compressed using motion compensation to provide a second compressed video signal. The data in the first and second compressed video signals is quantified. A comparison is made to determine which of the signals contains the least data. Successive sets of pixel data are sequentially compressed and quantified and the compressed video signal having the least data for each particular set is selected. The selected signals are encoded to identify them as motion compensated or non-motion compensated signals, and combined to provide a compressed video signal data stream for transmission. Apparatus for receiving and decoding the signals is also disclosed.

Identification of spatially quantised tachistoscopic images of faces: How many pixels does it take to carry identity?

Abstract: Six images of human faces were quantised into isoluminant square-shaped pixels (16 grey levels) at eight different spatial levels of quantisation The subjects had to identify the faces that were presented with different exposure durations (from 1 to 200 msec) and with one of two brightness conditions (variable brightness in Experiment 1 or isobrightness in Experiment 2) All finer quantisation levels led to better identification than the most coarse quantisation level (15 pixels per face in the horizontal dimension) at all exposure durations The observation of an abrupt decrease in identification efficiency on moving from 18 or more pixels per face to 15 pixels per face and the approximate equality in identification efficiency within a broad range of quantisation levels above 18 pixels per face pose some problems for existing theories of face recognition The implications of these findings for prototype-related, auto correlation and micro genetic accounts of face and pattern processing are disc

Layered DCT video coder for packet switched ATM networks

Abstract: A digital video encoder encodes a video frame into a differential video frame for transmission over a packet switched network. The video encoder includes an inter-frame encoder, an intra-frame encoder, and an encoding selector for selecting between the inter-frame and intra-frame encoder depending on the relative motion between the video frame being encoded and the previous video frame. A composite frame combiner provides a composite intra/inter-frame encoded difference frame having one set of pixels encoded by the inter-frame encoder, and another set of pixels encoded by the intra-frame encoder. The set of intra-frame encoded pixels includes at least one square or rectangular pixel block, a vertical strip of pixel blocks, and a horizontal strip of pixel blocks. Difference frames are encoded into separable data sets representing video information within a particular range of image resolution. A discrete cosine transform (DCT) is used to transform the difference pixels into corresponding DCT coefficients which are separable, by resolution, into the data sets providing coefficient layers. A packetizer formats the data sets into asynchronous transfer mode (ATM) packets for transmission over network.

Method and apparatus for rendering anti-aliased polygons

Abstract: A scan conversion process is performed on a polygon using a single pass technique. The pixels which comprise the edges and vertices of the polygon are first determined from the vertices which define the polygon. The alpha channel comprises either a sub-pixel mask associated with each pixel which indicates the amount and sub-pixel regions of coverage or a single value indicative of the percentage of coverage of a pixel. Furthermore, a z value indicative of the depth of each pixel is maintained. The pixels between the edge pixels of the polygon are then turned on, thereby filling the polygon. The pixels which comprise the polygon are then composited with the background pixels on a per pixel basis. The depth value of each pixel of the polygon (the z value) is used to determine the compositing equations to be used to composite each pixel of the polygon to the background. The compositing equations update the color of the pixel, the z buffer value of the background pixel and the sub-pixel mask to reflect the addition of information from the compositing of the pixel of the polygon into the background pixel. Through this method high quality anti-aliased polygons may be rendered without performing the time consuming process of sorting the polygons in depth order prior to compositing.

Elimination of systematic error in subpixel accuracy centroid estimation [also Letter 34(11)3347-3348(Nov1995)]

Abstract: An analysis of the properties of the centroid method for subpixel accuracy image feature location is presented. This method is free of systematic error if the maximum spatial frequency of the image incident on the image sensor is less than the sensor's sampling frequency. This can be achieved by using a lens aperture setting such that the modulation transfer function cut-off frequency due to diffraction is appropriately small. Both simulation and experimental tests of this prediction are presented for the case of the location of the center lines of the images of projected light stripes in a triangulation-based three-dimensional shape measurement system.

Hybrid coding system for moving image

Abstract: A moving image is coded through a waveform coding predictor (2), and a model-based coding predictor (3) which is mainly applicable to a head portion of a person, the coded outputs of said two predictors are compared (5) with an input current image for each block element which has for instance 8×8 or 16×16 pixels. A block element coded for the smaller difference from an input current image is taken, and is transmitted to a receiving side through differential coding (6,7). As the information compression ratio is high, a moving image can be transmitted through conventional telephone line, or a narrow band circuit.

Method and apparatus for diagnosis of breast tumors

Abstract: An apparatus for distinguishing benign from malignant tumors in ultrasonic images of candidate tissue taken from a patient. A region of interest is located and defined on the ultrasonic image, including substantially all of the candidate tissue and excluding substantially all the normal tissue. The region of interest is digitized, generating an array of pixels intensity values. A first features is generated from the arrays of pixels corresponding to the angular second moment of the pixel intensity values. A second feature is generated from the array of pixels corresponding to the inverse contrast of the pixel intensity values. A third feature is generated from the array of pixels corresponding to the short run emphasis of the pixel intensity values. The first, second and third feature values are provided to a neural network. A set of trained weights are applied to the feature values, which generates a network output between 0 and 1, whereby the output values tend toward 1 when the candidate tissue is malignant and the output values tend toward 0 when the candidate tissue is benign.

Gain calibrating nonuniform image-array data using only the image data

Abstract: An algorithm is developed for calibrating the spatial nonuniformity of image-array (CCD-type) detectors. Like other techniques this approach uses multiple, spatially displaced images. In circumstances where high-precision flat fields are not available by other means (i.e., sky flats) this technique is advantageous as it uses the data frames for gain calibration even when the array images extended, nonuniform, sources. Numerical experiments and direct observations with intrinsically uniform and quite nonuniform detectors show that this algorithm is useful when data frames are crowded with sources - circumstance where 'median filtering' flatfielding techniques often fail. The algorithm described is robust and efficiently uses information from multiple data frames to determine pixel gain variations, using visible and IR array observations of extended sources.

Tile-oriented technique for collectively performing image rotation, scaling and digital halftone screening

Abstract: A tile-oriented technique and associated apparatus for manipulating a continuous tone (contone) image through image rotation, anamorphic scaling and digital halftone screening for use in illustratively implementing a page description language. Specifically, an incoming contone image is first partitioned into aligned non-abutting tiles. Overlapping blocks are then defined which will hold output data for corresponding tiles. To effect rotation and anamorphic scaling of the contone image, two-dimensional sampling increments, in fast and slow scan directions, are defined to relate movement between successive pixels in an output block to movement between corresponding pixels in the contone image. Similar, though independent, sampling increments, based in part upon screen angle and screen ruling, are defined for movement between successive pixels in a halftone reference cell. Incremental sampling occurs in the contone image to yield a corresponding sampled contone value. This value, in conjunction with incremental halftone sampling addresses, then defines a sampling location that is read in a halftone reference plane, the resulting output of which is single bit halftone data that defines a writing spot. Each tile in the contone image is successively processed with resulting output data for that tile being written into appropriate pixel locations in a corresponding block in the output image. Clipping variables, incrementally varying in two-dimensional fashion and in unison with the contone pixel sampling location, define valid output data for a contone tile that is to be written into a corresponding block.

Video image processing apparatus including convolution filter means to process pixels of a video image by a set of parameter coefficients

Abstract: To process pixels included in a video image, control data is generated to established a pixel processing operation for a pixel under consideration. A set of parameter coefficients is generated in response to this control data; and a convolution filter receives successive pixels of an input video image for convoluting the pixel under consideration as well as surrounding pixels with the set of parameter coefficients, thereby producing a processed pixel which is used by further apparatus in place of the pixel under consideration.

Apparatus for detecting text edges in digital image processing

Abstract: Apparatus for detecting edges of black text in an original image. An edge detector which low pass filters the image signal to blur and widen any edges in the original image, and then high pass filters the low pass filtered signal to produce a first output signal which represents the blurred edges in the original image. The non-edge blurred signal is high pass filtered to produce a second output signal which represents the non-blurred edges in the original image. The first and second output signals are combined to create a first edge control signal that encompasses any edges in the original image. Halftone areas of the original image are identified by detecting dots in the original image, a second edge control signal that excludes areas in the original image identified as containing dots is created. Color pixels in the original image are detected, and a neutral color signal that represents the non-color portions of the original image is created. The first edge control signal, the second edge control signal, and the neutral color signal are combined to create a black text edge control signal that encompasses edges in the original image not in an area that contains dots.

Method and device for modifying a zone in successive images

Abstract: The process makes it possible to modify the content of a sequence of images which can be represented as a pixel matrix. For that a non-deformable target zone to be replaced in the images is identified and is characterized, both in position and in dimensions, by reference to marks located close to each other and having an unchanging geometrical relation with the target zone in each image. The position, with respect to the marks, to be assigned to a model or pattern which can be superimposed on the target zone, stored in digital form, is determined. For each particular image of a same shot the geometrical transformation to be carried out on the model so that it corresponds to variations in the appearance of the target zone in all the successive images in the same shot is determined automatically by a correlation based on the marks. Last the pixels representative of the model are substituted for the pixels representative of the target zone in the image.

Method and apparatus for storing and merging multiple optically scanned images

Abstract: A method for storing multiple scanned images of a response-marked form document comprises scanning a copy of the form document on which no response information has been marked, using a scanning method that detects substantially all visual information present on the form document and produces a pixel image thereof and storing that pixel image data in a computer storage device. Additional steps comprise scanning at least one copy of the same form document that has been marked with at least one response mark that absorbs light falling outside the visible spectrum, using a scanning method that detects only the presence of light outside the visible spectrum and that produces a pixel image of said at least one response mark, then compressing that pixel image to reduce the amount of pixel data used to represent areas of the form document bearing no response marks and storing the compressed pixel image data in a computer storage device. The visual scan pixel image and the image resulting from a scan for response marks can be merged together to produce a signal image that is comparable in quality and appearance to an image from a single, conventional visual-mode scan.

A new simple and efficient antialiasing with subpixel masks

Abstract: Antialiasing of edges is often performed with the help of subpixel masks that indicate which parts of the pixel are covered by the object that has to be drawn. For this purpose, subpixel masks have to be generated during the scan conversion of an image. This paper introduces a new algorithm for creating subpixel masks that avoids some problems of traditional algorithms, like aliasing of high frequencies or blinking of small moving objects. The new algorithms can be implemented by lookup tables that make use of the inherent symmetry of the algorithm. The results are compared with conventional supersampling. A hardware implementation is described.

Image skeletonization method

Abstract: A method for improved thinning or skeletonizing handwritten characters or other variable-line-width images. The method scans a template set over the image to be thinned. Each template has a specific arrangement of dark and light pixels. At least one of those templates includes either more than three pixels per row or more than three rows of pixels. An odd number is good choice. Moreover, the templates are chosen so that each template can unconditionally delete image pixels without consideration of the effect of such deletions on the behavior of the other templates. Thus the templates are independent of each other.

Electrophoretic display employing grey scale capability utilizing area modulation

Abstract: An apparatus provides grey scale capability for an electrophoretic information display (10). The electrophoretic information display (10) is an XY addressable display with each XY coordinate indicative of a given column (23) and row (25) and defining a pixel. There are described means coupled to display for impressing upon the display a predetermined digital pattern to cause certain pixels in the display to be energized with respect to other pixels in the display where the combination due to the digital pattern causes a desired grey scale level to appear on a character or on the background of the display.

Method for maintaining bit density while converting images in scale or resolution

Abstract: A method is provided for increasing resolution of an input image in an input bitmap while performing a density preserving smoothing operation. The input bitmap includes a first set of pixels K×L×b, and the method comprises the following steps: (a) mapping the first set of pixels into a second set of individual pixels κK× λL×b, each of the individual pixels of the second set having a value selected from a set of "b" output states; (b) developing pixel arrays from the second set of individual pixels, the pixel arrays respectively representing local neighborhoods for the individual pixels of the second set, each of the local neighborhoods having a value selected from a set of "c" output states; (c) combining the output state of each individual pixel with the output state of its respective local neighborhood resulting in a selected one of a plurality of decision states, each of the decision states reflecting a current output state of a selected individual pixel and preference of the selected individual pixel for one of the selected b output states, at least one of the decision states having at least one counterpart in the set of decision states; (d) examining the individual pixels within a preselected area to locate a preselected number of individual pixels including a first individual pixel having a selected one of the decision states and at least a second individual pixel having a counterpart value to the selected one of the decision state of the first individual pixel; and (e) altering the respective output states of the first and second individual pixels when each of the first and second individual pixels are within the preselected area wherein the resolution of the output image is changed and the output image is smoothed without altering the density of the output image relative to the input image.

Method for inspecting components having complex geometric shapes

Abstract: A method for inspecting a component of a gas turbine engine or the like having a plurality of similarly shaped structural portions, such as the gear teeth of a gear, the dovetail slots of a turbine disk or the like, includes the steps of: scanning a surface of at least one of the similarly shaped structural portions with an eddy current probe to induce eddy currents in the component; generating a two-dimensional image of the at least one portion from eddy current signals received during scanning, the image including a multiplicity of pixels arranged in a two-dimensional array and each pixel having a gray scale intensity responsive to the eddy current induced in the component at a component location corresponding to a position of the pixel in the matrix array; preprocessing the image to substantially reduce any signals or changes in the gray scale intensity of any pixels relative to the background pixel intensities of the image caused by geometrical characteristics and background noise common to all similarly shaped structuaral portions; identifying any suspected defect regions from the preprocessed image; determining a defect signal for each suspected defect region; and rejecting the component if any defect signal exceeds a predetermined reference value.

Method and apparatus for assembling a composite image from a plurality of data types

Abstract: A method and apparatus for combining multiple image data files of differing sizes, resolutions, and formats in real time into a single data stream for conversion into halftone pixel data for printing by either single or multiple beam recording devices The data of varying formats is received via a host interface which couples the data to a standard VME bus The internal CPU reformats the data and stores it on disk memory The data is routed by the CPU to one of a plurality of scan-line video processors which converts the data into halftone pixel data The halftone data is routed to a single or multiple beam recording device for printing A special low resolution image is prepared by the CPU for preview at a video work station

Color image processing system for preparing a composite image transformation module for performing a plurality of selected image transformations

Abstract: A method and apparatus for generating an array of modified pixel values in response to an array of input pixel values. The apparatus includes a user interface for receiving a user's selection of an image transformation(s) to be performed on the array of input pixel values. A storage device stores a plurality of transform definitions, each transform definition including sample values representing an input/output relation of an image transformation. A transform controller selects a plurality of transform definitions in response to the user's selection of at least one image transformation, and generates therefrom a composite transform definition. The composite transform definition includes sample values of an input/output relation of a composite image transformation which is equivalent to the image transformation(s) selected by the user. The transform definitions stored by the storage device include a plurality of predetermined transform definitions. A custom transform definition can be prepared in accordance with a user's instructions and added to the plurality of transform definitions stored by the storage device. The transform controller operates in essentially real time, implementing the user's selections sufficiently quickly to allow the user to interact with the image processing system until a desired array of modified pixel values are obtained.

Rectilinear object image matcher

Abstract: A process for extracting the precise perimeter of buildings from down-looking terrain images. The process works for any rectilinear building shape, and no restrictions are placed on the orientation or scale of the buildings within an image. The process has the advantages of robustness and execution speed. The process is robust in that it can extract building perimeters from a digitized image in the presence of multiple noise sources. The execution speed is proportional to N+M, where the size of the image is N by M pixels.

Unsharp masking using center weighted local variance for image sharpening and noise suppression

Abstract: A pseudo center weighted local variance in the neighborhood of an image pixel determines the amplification factor multiplying the difference between the image pixel and its blurred counterpart before it is combined with the original image. The amplification factor is computed for each pixel in the image, and varies from a minimum value of about -1 when the center weighted variance is at a minimum value to some maximum value when the center weighted variance reaches a maximum value. The pseudo center weighted variance is computed by convolving the local neighborhood surrounding each image pixel with a kernel of center weights having a maximum value at the center pixel. The pseudo center weighted variance is the difference between the center weighted mean of the source of the local neighborhood of image pixels and the square of the center weighted mean. Both image sharpening and noise suppression is achieved in the same image by permitting the amplification factor to vary between a fraction of unity and greater than unity as the center weighted variance varies from a low value (uniform image) to a higher value (textured image).