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 an image; a first convolution operator for generating a feature map by performing a convolution operation on the basis of the image; and a compressor for compressing the feature map into units of at least one line; and a decompressor for reconstructing the feature map compressed into units of lines.

Abstract: A semiconductor device includes an image recognition device having a convolution arithmetic processing circuit. The convolution arithmetic processing circuit includes a coefficient register where coefficients of an integration coefficient table are set, a product calculation circuit that calculates products of an input image and the coefficients, a channel register where a channel number of the integration coefficient table is set, a channel selection circuit that selects an output destination of a cumulative addition arithmetic operation on the basis of the channel number, and a plurality of output registers that store a result of the cumulative addition arithmetic operation. The integration coefficient table is a table where a plurality of input coefficient tables are integrated and the integration coefficient table has a size of N×N. The product calculation circuit can calculate data of N×N all at once.

Abstract: A signal to noise ratio adjustment circuit is configured to determine, whether a signal to noise ratio of a first image is below a first threshold and to determine, whether a variation of imaged content between the first image and a preceding second image of a video sequence is below a second threshold. The signal to noise ratio adjustment circuit is further configured to generate a third image having an increased signal to noise ratio as compared to the first image or the second image if the signal to noise ratio is below the first threshold and if the variation is below the second threshold.

Abstract: Disclosed is an electronic apparatus which executes an artificial intelligence algorithm using a convolution calculation unit and memory local to the electronic apparatus.

Abstract: Neural network processors including a window expander circuit and related methods are provided. The window expander circuit may include a first logic circuit configured to store a set of data elements, corresponding to at least a subset of the input data, into a Q number of logical memories, where each of a P number of data elements of the set of the data elements is stored in each of the Q number of logical memories. The window expander circuit may further include a second logic circuit configured to receive the first set of data elements and additional data elements corresponding to the at least the subset of the input data from the Q number of logical memories and expand the at least the subset of the input data until the at least the subset of the input data is expanded based on a predetermined factor.

Abstract: A method and system for improving data storage, presentation, and manipulation is provided. The method includes receiving data from a data source and analyzing the data for data classification attributes. A data set associated with a data classification attribute is selected and a data texture characteristic for association with said data set is received. The data texture characteristic is associated with the data set and a data texture representation associated with the data texture characteristic is presented via a haptic hardware device. A touch-based command associated with the data set is received and a user is enabled to manipulate, transfer, and store the data set.

Abstract: A processor-implemented neural network method includes: obtaining, from a memory, data an input feature map and kernels having a binary-weight, wherein the kernels are to be processed in a layer of a neural network; decomposing each of the kernels into a first type sub-kernel reconstructed with weights of a same sign, and a second type sub-kernel for correcting a difference between a respective kernel, among the kernels, and the first type sub-kernel; performing a convolution operation by using the input feature map and the first type sub-kernels and the second type sub-kernels decomposed from each of the kernels; and obtaining an output feature map by combining results of the convolution operation.

Abstract: Disclosed is a system for operating a network incorporating high-density connections for increased efficiency, network operation and management. The high-density connections are incorporated into the patch panel, network switch, and cables that connects them, as well as into cable analyzers and printed circuit boards (PCBs) which allow for a complete network within a single computer running virtualization software.

Abstract: Embodiments of the present disclosure relate to a configurable convolution engine that receives configuration information to perform convolution or its variant operations on streaming input data of various formats. To process streaming input data, input data of multiple channels are received and stored in an input buffer circuit in an interleaved manner. Data values of the interleaved input data are retrieved and forwarded to multiplier circuits where multiplication with a corresponding filter element of a kernel is performed. Varying number of kernels with different sizes and sparsity can also be used for the convolution operations.

Abstract: Systems, methods, and apparatus, including computer programs encoded on a computer storage medium. In one aspect, a system includes initial neural network layers configured to: receive an input image, and process the input image to generate a plurality of first feature maps that characterize the input image; a location generating convolutional neural network layer configured to perform a convolution on the representation of the first plurality of feature maps to generate data defining a respective location of each of a predetermined number of bounding boxes in the input image, wherein each bounding box identifies a respective first region of the input image; and a confidence score generating convolutional neural network layer configured to perform a convolution on the representation of the first plurality of feature maps to generate a confidence score for each of the predetermined number of bounding boxes in the input image.

Abstract: An apparatus is described. The apparatus includes an image processing unit. The image processing unit includes a plurality of stencil processor circuits each comprising an array of execution unit lanes coupled to a two-dimensional shift register array structure to simultaneously process multiple overlapping stencils through execution of program code. The image processing unit includes a plurality of sheet generators respectively coupled between the plurality of stencil processors and the network. The sheet generators are to parse input line groups of image data into input sheets of image data for processing by the stencil processors, and, to form output line groups of image data from output sheets of image data received from the stencil processors. The image processing unit includes a plurality of line buffer units coupled to the network to pass line groups in a direction from producing stencil processors to consuming stencil processors to implement an overall program flow.

Abstract: In one embodiment of the present invention a convolution engine configures a parallel processing pipeline to perform multi-convolution operations. More specifically, the convolution engine configures the parallel processing pipeline to independently generate and process individual image tiles. In operation, for each image tile, the pipeline calculates source locations included in an input image batch. Notably, the source locations reflect the contribution of the image tile to an output tile of an output matrix—the result of the multi-convolution operation. Subsequently, the pipeline copies data from the source locations to the image tile. Similarly, the pipeline copies data from a filter stack to a filter tile. The pipeline then performs matrix multiplication operations between the image tile and the filter tile to generate data included in the corresponding output tile. To optimize both on-chip memory usage and execution time, the pipeline creates each image tile in on-chip memory as-needed.

Abstract: A method is described. The method includes repeatedly loading a next sheet of image data from a first location of a memory into a two dimensional shift register array. The memory is locally coupled to the two-dimensional shift register array and an execution lane array having a smaller dimension than the two-dimensional shift register array along at least one array axis. The loaded next sheet of image data keeps within an image area of the two-dimensional shift register array. The method also includes repeatedly determining output values for the next sheet of image data through execution of program code instructions along respective lanes of the execution lane array, wherein, a stencil size used in determining the output values encompasses only pixels that reside within the two-dimensional shift register array.

Abstract: Provided is an image-acquisition apparatus that includes: a stage on which a specimen is mounted; an objective lens that collects light from the specimen mounted on the stage; a motor that drives the stage in the direction intersecting an optical axis of the objective lens; an imaging device that acquires images by capturing the light collected by the objective lens; and a processor comprising hardware, wherein the processor is configured to implement: a deterioration-level calculation unit configured to calculate deterioration levels contained in the acquired images; and an image generating unit configured to generate a pasted image by pasting the acquired images, wherein the image generating unit is configured to generate the pasted image by combining, for a common region of two images that are mutually pasted, an image with lower deterioration level calculated by the deterioration-level calculation unit at a higher combining ratio.

Abstract: Digital integrated circuit (IC) for extracting features out of input image is disclosed. The IC contains one or more identical cellular neural networks (CNN) processing engines operatively coupled to at least one I/O data bus. Each CNN processing engine includes a CNN processing block, a first set of memory buffers for storing imagery data and a second set of memory buffers for storing filter coefficients. CNN processing block is configured to simultaneously perform 3×3 convolutions at M×M pixel locations using received imagery data and corresponding filter coefficients. Imagery data represents a (M+2)-pixel by (M+2)-pixel region of the input image. CNN processing block further performs rectification and/or 2×2 pooling operations as directed. When two or more CNN processing engines are configured on the IC, CNN processing engines are connected to one another as a loop via a clock-skew circuit for cyclic data access. M is a positive integer.

Abstract: The present invention relates to a method, a system, and a light source for penetrating radiation imaging, and more particularly, to a method, a system, and a light source for X-ray imaging. The system for X-ray phase contrast and high resolution imaging of the present invention comprises an X-ray source comprising a plurality of X-ray micro-light sources, an X-ray sensor configured to receive X-rays penetrating an object, and a computer configured to receive and compute raw image data from the X-ray sensor so as to obtain a clear image of the object.

Abstract: Phase variations of the transverse magnetization in magnetic resonance induced by superimposed physical phenomenae or by intrinsic deviations of the main magnetic B0 field are separated from Feature Space set by demodulation and deconvolution, either by electrical circuits or by equivalent computational methods, permitting mapping and measurement of these induced phase variations independent of Feature Space.

Abstract: A convolution circuit includes: a plurality of input oscillators, each configured to: receive a corresponding analog input signal of a plurality of analog input signals; and output a corresponding spiking signal of a plurality of spiking signals, the corresponding spiking signal having a spiking rate in accordance with a magnitude of the corresponding analog input signal; a plurality of 1-bit DACs, each of the 1-bit DACs being configured to: receive the corresponding spiking signal of the plurality of spiking signals from a corresponding one of the input oscillators; and receive a corresponding weight of a convolution kernel comprising a plurality of weights; output a corresponding weighted output of a plurality of weighted outputs in accordance with the corresponding spiking signal and the corresponding weight; and an output oscillator configured to generate an output spike signal in accordance with the plurality of weighted outputs from the plurality of 1-bit DACs.

Abstract: A user position can be detected or determined using a sensor. A transformation can be applied to the image data based on the user position. The transformed image data can be displayed on a display.

Abstract: A retrieval support system includes a database which stores commodity image data; a retrieval unit which acquires commodity image data having image characteristic information which is the same as or similar to image characteristic information indicating a characteristic in an image of input image data, from the database, with respect to the input image data; and a condition setting unit which sets a retrieval condition including at least one of time information relating to time, place information relating to places and occasion information relating to occasions. The retrieval unit performs retrieval of the commodity image data based on the retrieval condition set by the condition setting unit. The commodity image data acquired by the retrieval unit and information relating to a commodity different from a commodity corresponding to the commodity image data are output together.

Abstract: A method and device for estimating the opacity at a point of a scene lit by an area light source and comprising an object defined by a mesh and occluding some of the emitted light. In order to optimize the calculations for live estimation of the opacity, the method comprises sampling said area light source in a plurality of samples, for at least one sample of the plurality of samples and for at least one first mesh element of the occluding object visible from the at least one sample, generating one shadow plane per edge of the at least one first mesh element, estimating a opacity level depending on coefficients of projection in a function base from values representative of the opacity for a set of intersection points between at least one ray having for origin a viewpoint of the scene and shadow planes crossed by said at least one ray, depending on an angle formed by the normal associated with each shadow plane crossed and by said at least one ray.

Abstract: A method of producing barcode locations within a document comprising providing an original digital image of the document; using a processor for producing a normalized image from the original digital image; producing a gradient image from the normalized image; producing a plurality of bounding boxes from the gradient image; and producing bounding box test results by testing each bounding box for the presence of a barcode; and producing barcode locations within the document from the bounding box test results.

Abstract: There are provided: an attribute determination unit configured to determine an attribute of an object included in image data to be subjected to printing processing; a color conversion unit configured to convert the color space of the image data into a color space compatible with a printing apparatus in accordance with the determination results obtained by the attribute determination unit; an edge determination unit configured to determine an edge of an object included in the image data whose color space has been converted by the color conversion unit; and a black monochromatization unit configured to convert the color of a pixel, which is a pixel of an achromatic object of the objects of the image data whose color space has been converted into the color space compatible with the printing apparatus and which is a pixel determined to be an edge by the edge determination unit, into black alone.

Abstract: A method comprises computing a color factor value indicating an amount of color gradients in at least one color channel from the query image. The method comprises combining the color-keypoints with the gray-keypoints when the color factor value is greater than a threshold. A method for performing a visual search comprises extracting a plurality of local descriptors from a query image and then selecting a subset of them based on various criteria's such as visual meaning score. A method comprises aggregating each mean vector for each visual codeword from distances between each visual codeword and local descriptors. The method comprises aggregating variance vector for each visual codeword from the distance between each visual codeword, and local descriptors. The method comprises transmitting aggregated mean vector information and aggregated variance vector information to a search server for efficient image retrieval.

Abstract: Described is a technology, such as implemented in a computational software program, by which a minimal polynomial is efficiently determined for a radical expression over the ring Z of integer numbers or the ring Q of rational numbers. The levels of the radical are grouped into a level permutation group that is used to find a level permutation set. An annihilation polynomial is found based upon the level permutation set. The annihilation polynomial is factored, and a selection mechanism selects the minimal polynomial based upon the annihilation polynomial's factors.

Abstract: A learning device includes a gradient feature extraction unit which extracts a gradient feature amount including a gradient direction at each coordinate and a gradient intensity value thereof based on an amount of variation between luminance at each coordinate of an inputted learning target pattern and luminance at a periphery thereof, a sum difference feature extraction unit which calculates a predetermined sum difference feature amount by adding the gradient intensity values according to the gradient directions included in a predetermined gradient range indicating a range of the predetermined gradient direction based on the extracted gradient feature amount and subtracting the gradient intensity values according to the gradient directions included in the other gradient range adjacent to the predetermined gradient range from the calculated sum, and a learning unit which acquires a learning parameter at each coordinate.

Abstract: Technologies are generally described for generating depth data based on a spatial light pattern. In some examples, a method of generating depth data includes obtaining an image of one or more objects on which a spatial light pattern is projected, wherein blurring of the spatial light pattern in the image monotonously increases or decreases in a depth direction, calculating a value of a spatial frequency component of the image in a local image area around a pixel of interest, and determining depth data corresponding to the calculated value of the spatial frequency component by utilizing a preset relationship between depths and values of the spatial frequency component.

Abstract: The present invention relates to the parallel calculation of convoluted data. In particular, the invention relates to Gaussian pyramid construction and parallel processing of image data, such as parallel calculation of repeatedly convoluted data for use in a SIFT algorithm.

Abstract: A convolution image processor includes a load and store unit, a shift register unit, and a mapping unit. The load and store unit is configured to load and store image pixel data and allow for unaligned access of the image pixel data. The shift register is configured to load and store at least a portion of the image pixel data from the load and store unit and concurrently provide access to each image pixel value in the portion of the image pixel data. The mapping unit is configured to generate a number of shifted versions of image pixel data and corresponding stencil data from the portion of the image pixel data, and concurrently perform one or more operations on each image pixel value in the shifted versions of the portion of the image pixel data and a corresponding stencil value in the corresponding stencil data.

Abstract: An apparatus and method for image processing of XYZ point clouds obtained from a GmAPD LADAR using low-pass filtering followed by high-pass filtering and deconvolution. Preferably, the low-pass filter parameters are developed numerically utilizing Weiner-Levinson Deconvolution (WLD).

Abstract: A method and apparatus for enhancing and improving image quality are provided. The method includes separating an input image into at least one low frequency component and at least one high frequency component; modulating the low frequency components in a block unit by dithering; modulating the high frequency components by sampling; and combining the modulated low frequency components and the modulated high frequency components.

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: A method of image processing includes receiving camera data representative of an object, implementing, with a processor, a convolution operation on the camera data to generate filtered data representative of the object, the convolution operation being configured with singular value decomposition (SVD) kernel data, and storing the filtered data in a memory. Implementing the convolution operation includes applying the SVD kernel data in parallel to a plurality of subsets of the camera data.

Abstract: In an image capturing and processing system, a device and method is provided. The lens is made of simple lens with high diffractive materials and known strong color aberrations. The method includes the steps of: calibrating or measuring a Point Spread Function (PSF) for each color components at a set of distances, capturing image data, and calculating the image obtained using a de-convolution method based on the PSF found.

Abstract: Disclosed are a system and a method for recognizing a disguised face using a Gabor feature and a support vector machine (SVM) classifier according to the present invention. The system for recognizing a disguised face includes: a graph generation means to generate a single standard face graph from a plurality of facial image samples; a support vector machine (SVM) learning means to determine an optimal classification plane for discriminating a disguised face from the plurality of facial image samples and disguised facial image samples; and a facial recognition means to determine whether an input facial image is disguised using the standard face graph and the optimal classification plane when the facial image to be recognized is input.

Abstract: A method of generating output image data comprises obtaining derivative data relating to a reference image; obtaining a constraint for output image data; and generating the output image data from the derivative data relating to the reference image in dependence on the constraint. This method can be used to recover a robust output image from the derivative of an input image.

Abstract: A computer implemented method for A computer implemented method for applying a numerical approximation of a convolution of image I as represented by hierarchical signals al of level l with filter f, said method including the steps of: generating a forward transformation by applying a convolution between al and kernel h1 for each level of hierarchy l and by down-sampling the result of convolved al and h1; generating a backward transformation by applying, for each level of hierarchy l, a convolution between kernel h2 and an up-sampled forward transformation and combining the result with a convolution of al with kernel g; and combining the forward transformation with the backward transformation, to yield â0 being an approximation of a convolution between a and f, wherein kernels h1, h1, and g are optimized kernels of filter f.

Abstract: A method for deconvolution of digital images includes obtaining a degraded image from a digital sensor, a processor accepting output from the digital sensor and recognizing a distorted element within the image. The distorted element is compared with a true shape of the element to produce a degrading function. The degrading function is deconvolved from at least a portion of the image to improve image quality of the image. A method of indirectly decoding a barcode includes obtaining an image of a barcode using an optical sensor in a mobile computing device, the image comprising barcode marks and a textual character. The textual character is optically recognized and an image degrading characteristic is identified from the textual character. Compensating for the image degrading characteristic renders previously undecodable barcode marks decodable. A system for deconvolution of digital images is also included.

Abstract: The present invention discloses a method for establishing evaluation standard parameters and method for evaluating the quality of a display image, wherein, the method comprises: taking pictures to a group of test images having different color shift severity degrees to obtain a sample picture group; selecting a standard picture by human eye; applying the Fourier transform to tristimulus values of all pictures; respectively applying convolution to the frequency distribution function corresponding to each primary color with a contrast sensitivity function of human eye; respectively normalizing to each of the convolution functions; and selecting the evaluation parameters of the three primary colors of the standard picture as the evaluation standard parameters. The present invention can obtain more objective and systemic evaluation standard parameters.

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: Images are prepared for blending with a reflection. Angles of incidence and angles or reflection lead to reflections of a viewer appearing to be different than actual size. Accordingly, image preparation can include scaling images of objects to be blended with a reflection to an appropriate size. For a flat specular surface the size is one half the size the object would be if the object were sized to be associated with the feature that is reflected. A viewer will focus on the reflection of the viewer at an object plane. An image associated with the specular surface will be blurred. Accordingly, the preparation can include compensating for the blurring. Preparation can also include reducing an area coverage of the image of the object to no more than about fifty percent of the region of the refection to provide space for the reflection.

Abstract: Embodiments related to the removal of blur from an image are disclosed. One disclosed embodiment provides a method of performing an iterative non-blind deconvolution of a blurred image to form an updated image. The method comprises downsampling the blurred image to form a blurred image pyramid comprising images of two or more different resolution scales, downsampling a blur kernel to form a blur kernel pyramid comprising kernels of two or more different sizes, and deconvoluting a selected image in the blurred image pyramid according to a Richardson-Lucy deconvolution process in which a bilateral range/spatial filter is employed.

Abstract: A system and method captures a moving image of a scene that can be more readily de-blurred as compared to images captured through other methods operating on equivalent exposure-time intervals. Rather than stopping and starting the light measurement during the exposure-time interval, photo-generated current is switched between multiple charge storage sites in accordance with a temporal switching pattern that optimizes the conditioning of the solution to the inverse blur transform. By switching the image intensity signal between storage sites all of the light energy available during the exposure-time interval is transduced to electronic charge and captured to form a temporally decomposed representation of the moving image.

Abstract: A system and method for displaying images of internal anatomy includes an image processing device configured to provide high resolution images of the surgical field from low resolution scans during the procedure. The image processing device digitally manipulates a previously-obtained high resolution baseline image to produce many representative images based on permutations of movement of the baseline image. During the procedure a representative image is selected having an acceptable degree of correlation to the new low resolution image. The selected representative image and the new image are merged to provide a higher resolution image of the surgical field. The image processing device is also configured to provide interactive movement of the displayed image based on movement of the imaging device, and to permit placement of annotations on the displayed image to facilitate communication between the radiology technician and the surgeon.

Abstract: A method of image processing includes receiving camera data representative of an object, implementing, with a processor, a convolution operation on the camera data to generate filtered data representative of the object, the convolution operation being configured with singular value decomposition (SVD) kernel data, and storing the filtered data in a memory. Implementing the convolution operation includes applying the SVD kernel data in parallel to a plurality of subsets of the camera data.

Abstract: An image processing method and an image processing system using a regionalized architecture are provided. In the method, a plurality of neighborhood pixels having at least one similarity characteristic in an image are classified into a group region having the same characteristics. Then, a unique group mark is assigned to each of the group regions, a regionalized chain code is used to describe an edge of each group region, such that subsequent image processing procedures can be applied to the neighborhood pixels of a currently processed pixel having same group mark. Therefore, the group regions can be recognized according to the group mark and regionalized chain code thereof, so as to simultaneously perform image processing procedures on those group regions.

Abstract: Various embodiments of methods and apparatus for motion deblurring in text images are disclosed. In one embodiment, a threshold-based text prediction for a blurred image is generated. A point spread function for the blurred image is estimated. A result of the threshold-based text prediction function is deconvolved based on the point spread function. The generating, estimating, and deconvolving are iterated at a plurality of scales, and a final deconvolution of a result of the iteratively deconvolving is executed.

Abstract: Various embodiments of methods and apparatus for image deblurring and sharpening using local patch self-similarity are disclosed. In some embodiments, an input blurred image is down-sampled to generate a downsized image. The downsized image is convolved with a blur kernel to obtain a smoothed image. For each of a plurality of patches of the input blurred image, a corresponding patch in the smoothed image is found. High frequency components between each of the plurality of corresponding patches in the smoothed image and corresponding patches of the downsized image are computed. The high frequency components are applied to the plurality of patches of the input blurred images to generate a deblurred version of the input blurred image.

Abstract: Embodiments of the present invention provide a system for performing image conversion operations. The system starts by receiving a request from a client for one or more pixel buffers containing a pixel-formatted, cropped, geometrically transformed, and/or color matched version of an image representation. The system then determines if a provider can provide the one or more pixel buffers. If so, the system calls the provider to generate the one or more pixel buffers containing the pixel-formatted, cropped, geometrically transformed, and/or color matched version of the image representation. Otherwise, the system calls the provider to generate one or more intermediate pixel buffers, generates a sequence of converters for converting the one or more intermediate pixel buffers, and calls the sequence of converters to generate the one or more pixel buffers containing the pixel-formatted, cropped, geometrically transformed, and/or color matched version of the image representation.

Abstract: Techniques for performing convolution filtering using hardware normally available in a graphics processor are described. Convolution filtering of an arbitrary H×W grid of pixels is achieved by partitioning the grid into smaller sections, performing computation for each section, and combining the intermediate results for all sections to obtain a final result. In one design, a command to perform convolution filtering on a grid of pixels with a kernel of coefficients is received, e.g., from a graphics application. The grid is partitioned into multiple sections, where each section may be 2×2 or smaller. Multiple instructions are generated for the multiple sections, with each instruction performing convolution computation on at least one pixel in one section. Each instruction may include pixel position information and applicable kernel coefficients. Instructions to combine the intermediate results from the multiple instructions are also generated.