Abstract: A convolution operation method and a processing device for performing the same are provided. The method is performed by a processing device. The processing device includes a main processing circuit and a plurality of basic processing circuits. The basic processing circuits are configured to perform convolution operation in parallel. The technical solutions disclosed by the present disclosure can provide short operation time and low energy consumption.

Abstract: A pooling operation method and a processing device for performing the same are provided. The pooling operation method may rearrange a dimension order of the input data before pooling is performed. The technical solutions provided by the present disclosure have the advantages of short operation time and low energy consumption.

Abstract: A convolution operation method and a processing device for performing the same are provided. The method is performed by a processing device. The processing device includes a main processing circuit and a plurality of basic processing circuits. The basic processing circuits are configured to perform convolution operation in parallel. The technical solutions disclosed by the present disclosure can provide short operation time and low energy consumption.

Abstract: A system for convolving an image includes a processing circuitry that retrieves the image including a set of rows, and a set of kernels, and merges serially all columns of each kernel, to generate a merged kernel. The processing circuitry executes parallelly multiple times, a multiply-accumulate (MAC) instruction on a row loaded in a corresponding vector register and a corresponding coefficient of the merged kernel and a load instruction on a subsequent row in one clock cycle. In the same clock cycle based on the MAC instruction, a logical shift operation is executed on the merged kernel to shift a current coefficient of the merged kernel with a subsequent coefficient such that the MAC instruction is executed on the subsequent row and the subsequent coefficient in the next clock cycle. Thus, each clock cycle is utilized by the system for executing both the MAC and load instructions.

Abstract: Systems and methods described herein provide for management of notifications. A computing device in communication with a client device receives a notification from at least one notification source. The notification may be for rendering on a screen at the client device. A policy engine of the computing device may identify a context of the notification. The policy engine may add a tag to the notification according to the identified context to control rendering of the notification at the client device. The computing device may transmit the notification with the tag to a notification service at the client device. The notification service may be configured to manage the notification at the client device in accordance with the tag.

Abstract: Disclosed are an image processing method and device using a line-wise operation. The image processing device, according to one embodiment, comprises: a receiver for receiving two-dimensional pixel information; at least one first line buffer for outputting the two-dimensional pixel information into a line-wise data line; an operator for performing a convolution operation on the basis of the data line data; and at least one second line buffer for outputting the data line on which the operation has been performed.

Abstract: The embodiments discussed herein involve flood filling a region with anti-aliasing. In forming a fill region, a candidate pixel can be included in the region based on a color of the pixel and also a color of a neighbor of the point. The inclusion basis may be a color distance between a seed color and the points, and a color distance between the seed color and the point's neighbor. Points in the region may be weighted according to their color distance relative to the seed color, where the color distance can also take into account alpha values. Flood filling may be anti-aliased by assigning alpha values to pixels in gaps between corners of the fill region, where an alpha value may be proportional to a point's contribution to the gap. Dimples in a fill region may be tested for and used to determine which of two flood fill algorithms to use.

Abstract: An image processing device includes one or more processors. The processors detects two or more first partial areas corresponding to each of two or more portions among a plurality of portions that are included in an object and that are set in advance from a first image, and detect two or more second partial areas corresponding to each of two or more portions among the portions from a second image. The processors extracts two or more first feature vectors from two or more of the first partial areas, and extract two or more second feature vectors from two or more of the second partial areas. The processors determines whether an object included in the first image and an object included in the second image are same, by using the first feature vectors and the second feature vectors.

Abstract: This disclosure relates to methods, non-transitory computer readable media, and systems that generate a virtual long exposure image from a sequence of short exposure images portraying a moving object. In various embodiments, the image transformation system aligns two digital images in the sequence of short exposure images. The image transformation system can determine a motion vector path for the moving object between the first digital image and the second digital image. The image transformation system can also blend pixels along the motion vector path to generate a blended image representative of the motion of the moving object between the first digital image and the second digital image. The image transformation system can generate additional blended images based on consecutive pairs of images in the sequence of digital images and generates a virtual long exposure image by combining the first blended image with the additional blended images.

Abstract: Embodiments disclosed herein may be useful for controlling access to data, and particularly to data intended for restricted access. In at least one embodiment, a system and method for dynamic data masking intercepts input/output (I/O) read requests for data in external memory, determines if the data to be read should be masked, and appropriately, dynamically masks data before the requested data is written into system memory. The system and method for dynamic data masking provides a technical improvement to computing systems by, for example, avoiding the need to create an entirely separate database with scrubbed data and, thus, resource and costs associated with creating a separate, scrubbed database.

Abstract: An apparatus and method of training a convolutional neural network (CNN) are provided. A method of training a CNN including a plurality of convolution layers stored in a memory involves approximating, using a processor, a convolution layer among the plurality of convolution layers using a low-rank approximation; reducing the number of output reconstruction filters of the approximated convolution layer; and modifying a structure of the CNN based on an approximation result and the reduced number of output reconstruction filters.

Abstract: A computer-implemented method and apparatus are described for deblurring an image. The method may include accessing the image that has at least one blurred region and, automatically, without user input, determining a first value for a first size for a blur kernel for the at least one blurred region. Thereafter, automatically, without user input, a second value for a second size for the blur kernel is determined for the at least one blurred region. A suggested size for the blur kernel is then determined based on the first value and the second value.

Abstract: Systems, methods, and articles of manufacture for automatic target recognition. A hypothesis about a target's classification, position and orientation relative to a LADAR sensor that generates range image data of a scene including the target is simulated and a synthetic range image is generated. The range image and synthetic range image are then electronically processed to determine whether the hypothesized model and position and orientation are correct. If the score is sufficiently high then the hypothesis is declared correct, otherwise a new hypothesis is formed according to a search strategy.

Abstract: There is provided an image processing device including a weight calculation unit that calculates a weight corresponding to each of a plurality of pixel values centering on a pixel of interest of an input image based on a feature amount calculated based on the plurality of pixel values centering on the pixel of interest, a regression coefficient reading unit that reads a regression coefficient stored for each class code determined based on a plurality of pixel values corresponding to the pixel of interest of the input image, and a pixel value calculation unit that calculates a pixel value of a pixel of interest of an output image by performing calculation using the plurality of pixel values, the weights, and the regression coefficients centering on the pixel of interest of the input image.

Abstract: Embodiments of the present invention provide for improved timing control in 2-D image processing to maintain a constant rate of fetches and pixel outputs even when the processing operations transition to a new line or frame of pixels. A one-to-one relationship between incoming pixel rate and outgoing pixel rate is maintained without additional clock cycles or memory bandwidth as an improved timing control according to the present invention takes advantage of idle memory bandwidth by pre-fetching a new column of pixel data in a first pixel block of a next line or frame while a new column of an edge pixel block on a current line is duplicated or zeroed out. As the edge pixel block(s) on the current line are processed, the data in the first pixel block of the next line or frame become ready for computation without extra clock cycles or extra memory bandwidth.

Abstract: Embodiments of the present invention provide for improved timing control in 2-D image processing to maintain a constant rate of memory fetches and pixel outputs even when the processing operations transition to a new line or frame of pixels. A one-to-one relationship between incoming pixel rate and outgoing pixel rate is maintained without additional clock cycles or memory bandwidth as an improved timing control according to the present invention takes advantage of idle memory bandwidth by pre-fetching a new column of pixel data in a first pixel block of a next line or frame while a new column of an edge pixel block on a current line is duplicated or zeroed out. As the edge pixel block(s) on the current line are processed, the data in the first pixel block of the next line or frame become ready for computation without extra clock cycles or extra memory bandwidth.

Abstract: In one embodiment, a method includes a computing device accessing a deconvolution function for a point spread function describing at least part of a user's eyesight. The computing device applies the deconvolution function to an image to be displayed on a display to the user. The computing device then displays on the display the image as deconvolved.

Abstract: A method of reducing blurring in an image of size greater than M columns by N rows of pixels, comprises deriving a blur kernel k representing the blur in the image, and deriving an inverse blur kernel k?1. The inverse blur kernel is given by (I) where h(m) is the sum of the first m terms of the series (II) ? is the Dirac delta, m is greater than 1, and h(m) is a two dimensional matrix of size M×N. The two dimensional matrix h(m) is convolved with the image over the whole image in the image pixel domain to produce an image with reduced blur. The method may be applied to a video sequence allowing the sequence of images to be deblurred in real time.

Abstract: Embodiments of the present invention provide methods and associated architecture of accuracy adaptive and scalable vector graphics rendering including rendering a graphic comprising a plurality of line segments by processing each of the plurality of line segments in a first pass, and processing each of a plurality of pixels through which the plurality of line segments pass in a second pass, automatically detecting one or more rendering errors of the graphic, and correcting the one or more rendering errors. Other embodiments may be described and/or claimed.

Abstract: A system and method are provided for approximating a diffusion profile utilizing gathered lighting information associated with an occluded portion of an object. In use, the present technique gathers information associated with an occluded portion of an object that is illuminated with a two-dimensional pattern of light including an edge which defines an illuminated portion and the occluded portion of the object. To this end, a diffusion profile of the object is approximated, utilizing such information.

Abstract: A method for selectively transforming a multi-dimensional input array comprising D dimensions includes, with a computing system, determining a D-dimensional convolution of the input array at only selected points in the array, each the convolution being a function of a product of D one-dimensional kernels; determining partial convolutions at each dimension iteratively, an iterative determination of one of the partial convolutions being determined, in part, from a previous iterative determination; and collecting transformed values from the convolutions into the input array to form a transformed input array.

Abstract: Embodiments of an apparatus, system and method are described for an edge-preserving smooth filter for low power architecture. A weighted pixel sum may be determined based on a weight of a central pixel and a weight of one or more neighboring pixels. The weight sum for the central pixel may be set to a power of two. An output of the central pixel may be displayed based on the weight sum and the weighted pixel sum. Other embodiments are described and claimed.

Abstract: Embodiments of the present invention provide for improved timing control in 2-D image processing to maintain a constant rate of memory fetches and pixel outputs even when the processing operations transition to a new line or frame of pixels. A one-to-one relationship between incoming pixel rate and outgoing pixel rate is maintained without additional clock cycles or memory bandwidth as an improved timing control according to the present invention takes advantage of idle memory bandwidth by pre-fetching a new column of pixel data in a first pixel block of a next line or frame while a new column of an edge pixel block on a current line is duplicated or zeroed out. As the edge pixel block(s) on the current line are processed, the data in the first pixel block of the next line or frame become ready for computation without extra clock cycles or extra memory bandwidth.

Abstract: Among other disclosed subject matter, a computer-implemented method for pattern matching includes receiving a pattern image, a mask image and a search image, the mask image having an arbitrary shape and identifying a portion of the pattern image. The method includes evaluating a normalized cross-correlation equation based on the pattern image, the mask image and the search image, including at least a convolution of the mask image and the search image. The method includes outputting a result of evaluating the normalized cross-correlation equation, the result indicating whether the search image matches the portion of the pattern image.

Abstract: A dual image source display system with an anti-aliased textual foreground and graphic image background, where display information from each source is combined, but only after the intensity level for each given pixel color component in the graphical image background is dimmed by an amount which is equal to the highest intensity level of any pixel color component in the same pixel as the given pixel color component.

Abstract: A method for rendering graphical data is provided. In one embodiment, the method includes rendering an aliased version of one or more polygons and sampling one or more edges of the aliased polygons. The method also includes calculating a curve that approximates the edge portion and intersects a set of pixels, determining the proportional areas of the pixels located between the curve and the aliased edge portion, and rendering an anti-aliased version of the edge portion based on the proportional areas. Various devices, machine-readable media, and other methods for anti-aliasing of a graphical object are also provided.

Abstract: An image processing apparatus includes a face-image detector configured to detect a region of a face image from an image supplied thereto. A face-direction detector is configured to detect a direction that a face in the detected face image is facing and a feature-position detector is configured to detect feature positions corresponding to features of the face from the detected face image and the detected face direction. A feature-value calculator is configured to calculate feature values at the detected feature positions and a mapper is configured to map the calculated feature values using predetermined mapping functions. A a recognizer is configured to recognize whether the detected face is a registered face, using the mapped feature values and feature values registered in advance.

Abstract: What is disclosed is a novel system and method for color trapping on halftoned bi-level bitmaps. Color edges are detected and edge pixels that need to be trapped are identified. The number of pixels qualified as edge pixels eligible for color trapping can be up to a pre-determined number of pixels away from the color edge. Estimates for the continuous-tone values are obtained for the dominant colors on each side of the two-color edge. The contone value of the dominant color on the opposing side of the two-color edge is assigned to the qualified edge pixels. Qualified edge pixels are re-halftoned using their assigned contone value so that halftones for one color are extended beyond the edge into the other color. The re-halftoned edge pixels are combined with the original bitmap to produce a new bitmap for the image. The new bitmap is then provided to an image output device.

Abstract: A digital image-processing device with a Bayer sensor and an image memory is provided in which the image data of the sensor is written into an image memory, and from this image memory, image data in the Bayer format with a length L and a width B is written continuously into a data buffer, and in which the sample values are combined by means of a computational device with the help of adders, in each case symmetrically to a central point of one or more (2n+1)×(2n+1) neighborhoods, and one or more (n+1)×(n+1) matrices are derived by means of the computational device, and from this (n+1)×(n+1) matrix or these matrices, with the help of additional adders, at least one n×n matrix is formed, and a first color component is in each case calculated from this by means of an adder network.

Abstract: A texture synthesis device, which has a determinator for determining a prioritization among regions to be synthesized of an area to be synthesized, as a function of image information in an area adjacent to the area to be synthesized, and a synthesizer for synthesizing the regions to be synthesized in an order that depends on the prioritization.

Abstract: Embodiments of the present invention provide for improved timing control in 2-D image processing to maintain a constant rate of memory fetches and pixel outputs even when the processing operations transition to a new line or frame of pixels. A one-to-one relationship between incoming pixel rate and outgoing pixel rate is maintained without additional clock cycles or memory bandwidth as an improved timing control according to the present invention takes advantage of idle memory bandwidth by pre-fetching a new column of pixel data in a first pixel block of a next line or frame while a new column of an edge pixel block on a current line is duplicated or zeroed out. As the edge pixel block(s) on the current line are processed, the data in the first pixel block of the next line or frame become ready for computation without extra clock cycles or extra memory bandwidth.

Abstract: An apparatus and method are provided for approximating a convolution function (e.g. a diffusion profile, etc.) utilizing a sum of Gaussian functions. In use, results of a plurality of Gaussian functions are calculated. The results of the Gaussian functions are further summed. To this end, an approximation of a convolution function is generated based on the sum. Further, an image is rendered utilizing the approximation.

Abstract: The invention relates to an image processing system for omnidirectional reading of optical patterns, for example of one-dimensional and two-dimensional codes. One object is to provide an image processing system which ensures high resolution with rapid processing at the same time in order, in particular, to allow processing in video real time and to provide reliable identification of the image information, in particular reliable and low-error edge detection, segmentation and object association. According to the invention, this is achieved by the use of a plurality of convolvers, which operate in parallel and/or using the time-division multiplexing method, for convolution of the digital image data, of a neighborhood processor (90) for reading and linking contour points (22), and/or of a statistical processor for reading a segment list.

Abstract: A method for superimposing optical information in a scanning microscope includes determining a transformation matrix, and superimposing first optical information of a CCD image and second optical information of at least one piece of second image information using the transformation matrix.

Abstract: A new transparency group may be rendered and blended with other, previously rendered, transparency groups, by using the Porter-Duff algebra available on the GPU even though the transparency groups include pre-multiplied color and alpha information. Additionally, the number of copies of the back buffer (the image information for the previously rendered transparency groups) required to properly render, blend and combine the new transparency group into the image information of previously rendered transparency groups may be minimized.

Abstract: A multi-chip system and method are disclosed that utilizes a plurality of graphics pipelines to perform large kernel convolution. Each graphics pipeline includes a standard rendering unit and a video data convolve unit. Each video data convolve unit receives video pixel data from the video output of the standard rendering unit. The video data convolve units are connected in a chain. Each group of one or more video data convolve units in the chain convolves the video pixel data received by the group. The last video data convolve unit in the chain outputs a stream of convolved pixels.

Abstract: Hybrid architecture of supersampling and computing distance from a feature edge or Bezier evaluation to address thin feature support in graphics systems. To avoid missing some features the technique creates a supersampling of a small number of supersamples to pick up the thin features. By supersampling, samples can be produced on both sides of a thin feature, which causes thin features to be detectable by some pixel. Now that the thin features hit some pixel, the quality is achieved by a distance-from-edge approach. For example, the technique can supersample four times in combination with the distance-from-edge approach, produce another four samples there resulting in a 16-sample result.

Abstract: A technique for performing an anti-aliasing operation by multiple graphics processing units includes utilizing a first graphics processing unit to generate a first subset of filtered data resulting from performing anti-aliasing processing and similarly utilize a second graphics processing unit to generate a second subset of filtered data. The first graphics processing unit then pulls a first portion of the second subset of filtered data from a first memory block of a temporary buffer and blends such pulled data with a first portion of the first subset of filtered data. Overlapping in time with the pulling and blending operation of the first graphics processing unit, the second graphics processing unit pulls a second portion of the first subset of filtered data from a second memory block of the temporary buffer and blends such pulled data with a second portion of the second set of filtered data.

Abstract: A plurality of relative orientation detection filters having orientation coefficients that represent a relative orientation of a connected sequence of pixel drawing regions with their pixel values distributed in a predetermined pattern in a buffer memory having the same structure as an image memory, and a plurality of smoothing filters having smoothing coefficients for corresponding relative orientations, are prepared. A rendering unit draws an image to the buffer memory. A relative orientation of an edge in this image is detected by means of selecting out one relative orientation detection filter that is most suitable for the relative orientation in question. Pixel values for the edge are smoothed by using the smoothing filter according to the detected relative orientation. Then, drawing data including the smoothed pixel values are written into the image memory.

Abstract: An apparatus for image reduction and a method thereof are used to reduce an original image, wherein the apparatus includes a weighting element, a memory element, a convolution element, and an adder. Moreover, the weighting element generates a plurality of effective weights according to an image reduction ratio. The memory element is used to store the original image. The convolution element connects to the weighting element and the memory element for performing the Convolution calculation. The adder connects to the convolution element for summing the results of the Convolution calculation so as to output a reduced image.

Abstract: Multiple display heads of a single graphics processor are exploited to perform antialiasing and other processing tasks. In one embodiment, two display heads of the same graphics processor are coupled to each other in a master/slave configuration via a pixel transfer path. The “master” display head receives pixels from the “slave” display head in addition to its own pixels, and pixel selection logic in the master display head can blend the two pixels or select either one to the exclusion of the other. If the two pixels correspond to different sampling locations in the same display pixel, the blended pixel is an antialiased pixel.

Abstract: A vehicle restraint system has a vision-based occupant classification system for control of airbag deployment during a crash scenario. The classification system utilizes two imaging sensors which together create a stream of paired images received and stored by an occupant classification controller. A computer program product of the controller utilizes the paired images to extract disparity/range features and stereo-vision differential edge density features. Moreover, the controller extracts wavelet features from one of the two paired images. All three features or maps are classified amongst preferably seven classifications by algorithms of the computer program product producing class confidence data fed to a sensor fusion engine of the controller for processing and output of an airbag control signal input into a restraint controller of the vehicle restraint system.

Abstract: Methods and apparatuses for anti-aliasing scan conversion. In one aspect of the invention, an exemplary method to scan convert an image on a data processing system includes: sampling the image in a first direction to generate first signals for points along a second line in a second direction using a closed form solution for a convolution integral with a first kernel; and weighting the first signals for the points according to a second kernel in the second direction to generate a second signal for a pixel. In one example according to this aspect, the closed form solution is tabulated in a look up table. After entries are looked up from the look up table according to the image along a first line in the first direction on a first point of the points, the entries are combined to generate one of the first signals for the first point.

Abstract: A multi-dimensional data enhancement system can be embedded within a video cable for providing enhanced video images from a video signal source to a display. The multi-dimensional data enhancement system can also be provided in a switching box that allows a user to connect multiple video signal sources to the box and then select a particular video signal source from the group. The selected signal is then enhanced by the switching box before the signal is shown on a display.

Abstract: A sensitive display system is described. The sensitive display system allows for mediation of content provided from a source to a recipient, such as from a web server to a web browser. Mediation may involve intercepting and augmenting the content so as to restrict or otherwise control information displayed on a display of the recipient device. In this way, the sensitive display system allows restricted information to be replaced with security status messages, or removed entirely. For example, when a user comes within a specified distance from the sensitive display area, a security classification level of the approaching user is used to determine whether there should be any change in the display. If so, a mediating device causes a browser push to occur, to thereby block or otherwise restrict some or all of the display.

Abstract: A mobile communication terminal comprises a camera unit (34) and a main CPU. In the camera unit (34), a processing system including a color conversion circuit (66) and an image compression circuit (68) generates high-resolution compressed still image data based on interpolated image data of a photographed object. Also, a processing system including an image decompression circuit (74), a color conversion circuit (72), and a thinning-out circuit (70) generates low-resolution decompressed still image data based on the high-resolution compressed still image data. The main CPU activates an image compression circuit (68) when a release key is operated, and activates an image decompression circuit (74) when generation of the compressed still image data is completed. The low-resolution decompressed image data generated by activation of the image decompression circuit (74) is obtained by the main CPU from the camera unit and written into a VRAM. As a result, a decompressed image is displayed on an LCD.

Abstract: A method (1005) generates one or more linear blends. Initially, the method has a list of already received one or more fill-paths defining a current linear blend, and a newly received fill-path, where the fill-paths each comprise a single colored parallelogram (eg. 700). The method adds (1090) the new fill-path to the list to become the last fill-path in the list, if the difference (1230) in color between the new fill-path and the last fill-path in the list is no greater, in each color channel, than a predetermined threshold value multiplied by the difference in color between the last and second-last fill-paths in the list. The predetermined threshold value is preset to such a value so that the new fill-path will not be added to the list if the new fill-path does not visually form part of the current linear blend. The method repeats this step for each new fill-path received until the aforementioned condition is not satisfied.

Abstract: An electro-optical device and a method for displaying an image are disclosed. A clear image with a clear profile can be displayed therein by processing input image data, for example input image data of TV broadcasting received by the device.

Abstract: A system and method for rendering a non-zero thickness line on a pixel-limited output device such that aliasing of the line is reduced. The edges defining a line segment are expanded to insure that any pixel touched by the line segment has its center included in the bounds of the line segment. The area of any pixel partially or fully covered by the expanded line is determined. If one edge of the line traverses the pixel, the area is determined according to whether a triangular or triangular plus parallelogram area is covered. If more than one edge of the line segment traverses a pixel, the area covered is computed based on the single edge case. With the area covered by the line segment known, the color or shading of the pixel is determined by linear interpolation between the line and the background.

Abstract: An n-filter according to the present invention includes a nonlinear filter, a pattern detector, and a switch. The nonlinear filter maintains a steep edge whose size is larger than a predetermined threshold in fluctuations of pixels constituting an input image signal and, at the same time, smoothes a non-edge portion that does not include the edge. The pattern detector detects a fine edge in the fluctuations of the pixels constituting the input image signal and notifies the switch that the fine edge exists. The switch outputs the input image signal or an image signal output from the nonlinear filter to the subsequent stage in accordance with the notification from the pattern detector.