Abstract: A computer-implemented method for explaining an image classifier, the method comprising: receiving an initial image, the initial image having been wrongly classified by the image classifier; receiving an initial gradient of a function executed by the image classifier generated while classifying the initial image, the function being indicative of a probability for the initial image to belong to an initial class; converting the initial image into a latent vector, the latent vector being a representation of the initial image in a latent space; generating a plurality of perturbation vectors using the initial gradient of the function executed by the image classifier; combining the latent vector with each one of the plurality of perturbation vectors, thereby obtaining a plurality of modified vectors; for each one of the plurality of modified vectors, reconstructing a respective image, thereby obtaining a plurality of reconstructed images; transmitting the reconstructed images to the image classifier; for each one o

Abstract: An example method includes obtaining, by processing circuitry, an image depicting a face of a person wearing eyewear and a reflection pattern generated by projecting light onto a lens of the eyewear and determining, by the processing circuitry, an optical parameter of the lens based on the image. The method further includes receiving, by the processing circuitry and from an electronic database, a personal characteristic data relating to a previously identified person and comparing, by the processing circuitry, the optical parameter of the lens to the personal characteristic data. The method further includes determining, by the processing circuitry and based on the comparison, a score indicative of a match between the person wearing eyewear and the previously identified person.

Abstract: A method comprises receiving a network of a plurality of nodes and a plurality of edges, each of the nodes comprising members representative of at least one subset of training data points, each of the edges connecting nodes that share at least one data point, grouping the data points into a plurality of groups, each data point being a member of at least one group, creating a first transformation data set, the first transformation data set including the training data set as well as a plurality of feature subsets associated with at least one group, values of a particular data point for a particular feature subset for a particular group being based on values of the particular data point if the particular data point is a member of the particular group, and applying a machine learning model to the first transformation data set to generate a prediction model.

Abstract: The present disclosure is directed to systems and methods for analyzing digital images to determine alphanumeric strings depicted in the digital images. An electronic device may generate a set of filtered images using a received digital image. The electronic device may also perform an optical character recognition (OCR) technique on the set of filtered images, and may filter out any of the set of filtered images according to a set of rules. The electronic device may further identify a set of common elements representative of the alphanumeric string depicted in the digital image, and determine a machine-encoded alphanumeric string based on the set of common elements.

Abstract: A computer-implemented method for analyzing a survivability of a structure is presented. A distance extending radially from the structure defining an area surrounding the structure may be determined. The area surrounding the structure may be scanned to identify surrounding objects. The area surrounding the structure may then be segmented to create a plurality of area segments. A density of surrounding objects in each of the plurality of area segments may be determined. A risk score for each of the plurality of area segments may then be determined based on the density of surrounding objects in each of the plurality of area segments. An overall wind exposure risk value may be provided for further use.

Abstract: A method for image compression and a circuit system thereof are provided. In the method, pixel values of an image are obtained. A compression scenario is decided, for example, a uniform-quantization manner or a non-uniform-quantization manner is used for an M-bit image being compressed to an N-bit image so as to decide codeword sections for the image. Every codeword section has a codeword distance. The codeword sections have a fixed codeword distance in the uniform-quantization manner. Alternatively, in the non-uniform-quantization manner, the image can be divided into multiple codeword sections having different codeword distances according to a brightness distribution. Afterwards, a random number is generated for deciding codeword and index for original value of each of the pixels. An index table is accordingly formed. The index table is provided for obtaining the codeword in a decoding process by querying a codebook with the index.

Abstract: Editing an image includes an image input unit reading displaying data of the image. The displaying data includes angular profiles of an object in the image, and the angular profiles include color information of the object corresponding to a plurality of viewing angles. Editing the image also includes a processor altering the angular profiles of the object to generate edited displaying data according to at least one editing instruction, and an image output unit outputting the edited displaying data.

Abstract: A system and a method for determining a gaze direction of a user viewing a scene is provided. The system comprises a camera for obtaining an image of at least one of the user's eyes, a depth information detection device for obtaining depth data related to the image, and a processing unit. The processing unit is configured to define a surface at a predetermined position relative to the user's eye based on the depth data, obtain a normalized image by projecting the image onto the surface and determine a normalized gaze direction based on the normalized image. The processing unit is further configured to determine a gaze direction based on the normalized gaze direction and the depth data.

Abstract: A computer-implemented method to capture and detect clusters in, or determined by, a set V of discrete digital data comprising; computing, from the set V, an abstract separation system ASS that consists of a finite set S, whose elements are called separations; of a predetermined transitive, antisymmetric and reflexive order relation ? on S; and of an order-reversing involution *: S?S, that is, a mapping s?s* with the property that, (s*)*=s and that r?s implies s*<r* for all r, s?S; predetermining a set of consistency requirements (CRs), that is, a set F of subsets of S; computing, from the ASS (S,?, *), one or more abstract tangles, that is, any set T?S that contains exactly one of each pair {s, s*} for s?S, and does not contain any of the forbidden configurations F?F as a subset; or determining that there is no abstract tangle; and determining that any abstract tangle T represents a cluster in, or determined by, the data set V.

Abstract: A method for identifying identity information includes: acquiring user data, the user data including identity information and account information of a user; establishing an association relationship between account information and identity information that are bound in the user data, and establishing an association relationship between two pieces of account information having common features in the user data; and determining a risk value of target identity information in the user data according to the established association relationships, and determining, according to the determined risk value, whether the target identity information has a risk of being used fraudulently.

Abstract: Provided is a method including obtaining a first data object including a first set of data entries, wherein each data entry of the first set of data entries includes text content associated with a time entry. The method includes generating a first data object score using the text content and the time entries included in the first set of data entries and using scoring parameters, determine that the first data object score satisfies a data object score condition; perform in response to the first data object score satisfying the data object score condition, a condition-specific action associated with the data object score condition.

Abstract: An encoder includes circuitry and memory coupled to the circuitry. The circuitry, in operation: generates (i) a first quantization matrix for transform coefficients included in a current block to be processed and (ii) a second quantization matrix for transform coefficients included in a low frequency domain among the transform coefficients included in the current block; and quantizes the transform coefficients included in the current block using at least one of the first quantization matrix or the second quantization matrix, in accordance with a size of the current block.

Abstract: Techniques of indexing a database and processing a query involve decomposing the residual term according to a projection matrix that is based on a given direction v. For example, for each database element of a partition, the residual for that database element is split into a component parallel to a given direction and a component perpendicular to that direction. The parallel component lies in a one-dimensional subspace spanned by the direction and may be efficiently quantized with a scalar quantization. The perpendicular component is quantized using multiscale quantization techniques. The quantized residual components and the center elements of each partition define the indexed database. Upon receipt of a query from a user, the inner products of q with the residual may be computed efficiently using the quantized residual components. From these inner products, the database elements that are most similar to the query are selected and returned to the user.

Abstract: Techniques for learned lossy image compression are described. A system may perform image compression using an image compression model that includes an encoder to compress an image and a decoder to reconstruct the image. The encoder and the decoder are trained using machine learning techniques. After training, the encoder can encode image data to generate compressed image data and the decoder can decode compressed image data to generate reconstructed image data.

Abstract: An image coding apparatus includes a block dividing unit configured to divide an input image into a plurality of blocks, a prediction unit configured to perform prediction based on coded pixels to generate prediction errors, a transformation unit configured to perform orthogonal transform to the prediction errors to generate transform coefficients, a quantization matrix generation unit configured to generate quantization matrices that are used to quantize the transform coefficients, a quantization matrix coding unit configured to calculate difference values by scanning the quantization matrices and to code the difference values, a quantization unit configured to generate quantization coefficients by quantizing the generated transform coefficients using the quantization matrices, and a coefficient coding unit configured to code the quantization coefficients, wherein the quantization matrix coding unit is configured to scan coefficients of the quantization matrices in a unidirectional manner to calculate the di

Abstract: A compression system includes an encoder and a decoder. The encoder can be deployed by a sender system to encode a tensor for transmission to a receiver system, and the decoder can be deployed by the receiver system to decode and reconstruct the encoded tensor. The encoder receives a tensor for compression. The encoder also receives a quantization mask and probability data associated with the tensor. Each element of the tensor is quantized using an alphabet size allocated to that element by the quantization mask data. The encoder compresses the tensor by entropy coding each element using the probability data and alphabet size associated with the element. The decoder receives the quantization mask data, the probability data, and the compressed tensor data. The quantization mask and probabilities are used to entropy decode and subsequently reconstruct the tensor.

Abstract: These present disclosure provides devices and techniques to compress a list of integers. A circuit may include a sorter to sort a list of integers and a subtractor to determine a range of integers represented in the list and to recursively subdivide the range into sub-ranges. The circuit may also include a bit determiner to determine an amount of information (e.g., bits) to use to add indications of the integers in the list to a compresses list information element based on the range and the sub-ranges.

Abstract: A storage device may program data differently for different memory areas of a memory. In some embodiments, the storage device may use different codebooks for different memory areas. In other embodiments, the storage device may modify bit orders differently for different memory areas. What codebook the storage device uses or what bit order modification the storage device performs for a particular memory area may depend on the bad storage locations specific to that memory area. Where different codebooks are used, optimal codebooks may be selected from a library, or codebooks may be modified based on the bad storage locations of the memory areas.

Abstract: A storage device may program data differently for different memory areas of a memory. In some embodiments, the storage device may use different codebooks for different memory areas. In other embodiments, the storage device may modify bit orders differently for different memory areas. What codebook the storage device uses or what bit order modification the storage device performs for a particular memory area may depend on the bad storage locations specific to that memory area. Where different codebooks are used, optimal codebooks may be selected from a library, or codebooks may be modified based on the bad storage locations of the memory areas.

Abstract: Identifying a subset of signifiers to analyze can include determining a set of distance metrics between a first signifier and each of a plurality of second signifiers, identifying a subset of the plurality of second signifiers to analyze based on the set of distance metrics using a computing device, and determining a relation between the subset of the plurality of second signifiers and the first signifier based a subset of the set of distance metrics.

Abstract: A method is provided in one example embodiment and includes establishing a communication channel between a first network element and a second network element; identifying a video processing characteristic for a data stream in a network environment; embedding a plurality of bits into the data stream in order to identify the video processing characteristic that has been identified by the first network element; and communicating the data stream with the plurality of bits to a next destination.

Abstract: A moving picture coding device is a moving picture coding device which codes a moving picture, and includes: a VBV modeling unit which simulates an occupancy amount of a coded data buffer (VBV buffer occupancy amount) at the time of decoding; and a bit rate control unit which determines a quantization width (quantization scale) according to a rate of change of the occupancy amount of the coded data buffer so that the occupancy amount of the coded data buffer increases.

Abstract: A display apparatus for post-processing the image of the contents, an image post-processor and a method for post-processing the image of the contents are disclosed. The display apparatus includes a communicator which communicates with a server, a display which displays contents received from the server, and a controller which determines an image post-processing method for elements of the contents based on source information of the contents and predetermined condition information, renders the elements which are post-processed according to the determined image post-processing method, and controls the display to display the contents generated after the rendering. Accordingly, the display apparatus can selectively implement image post-processing for the elements of the contents.

Abstract: Methods and network devices for communicating data packets in a set of transmissions between a transmitter and a receiver of a communications network are described. The method includes prior exchanging transmitter-provided packet information and receiver-provided packet information between the transmitter and the receiver, generating an initial compression dictionary before the first transmission, based on the exchanged transmitter-provided packet information and the exchanged receiver-provided packet information. At the transmitter, the method further includes compressing the data packets of the first and subsequent transmissions of the set using the initial compression dictionary, and sending the compressed data packets over the communications network to the receiver.

Abstract: The invention proposes modification of quantized coefficients for signalling of a post-processing method. Therefore, it is proposed a method for lossy compress-encoding data comprising at least one of image data and audio data. Said method comprises determining quantized coefficients using a quantization of a discrete cosine transformed residual of a prediction of said data. Said method further comprises modifying said quantized coefficients for minimizing rate-distortion cost wherein distortion is determined using a post-processed reconstruction of the data, the post-processed reconstruction being post-processed according to a postprocessing method, and compress-encoding said modified coefficients. In said proposed method, the post-processing method is that one of n>1 different predetermined post processing method candidates whose position in an predetermined order of arrangement of the post processing method candidates equals a remainder of division, by n, of a sum of the modified coefficients.

Abstract: A system and method for compressing digital image data. An image may be divided into sub-regions. A maximum number of template codes may be determined for representing each sub-region, each template code uniquely identifying a template pattern in a dictionary. Each of a plurality of the sub-regions may be associated with a plurality of template patterns, each of which individually does not match the sub-region, but which when combined together more closely match the sub-region. A compressed data set may be generated for the image, in which each sub-region is represented by up to the maximum number of template codes uniquely identifying the associated plurality of template patterns. The up to the maximum number of template codes may be transmitted for each sub-region from the compressed data set, wherein the maximum number of template codes is a dynamic parameter adjustable based on the bandwidth available for transmitting.

Abstract: A device includes a coder or a codec configured for interleaved image data utilizing diamond shaped blocks for motion estimation and/or motion compensation and utilizing square or orthogonal transforms of residual data. In various embodiments, the decoder may be configured, among others, to perform de-blocking on edges of the diamond shaped blocks and/or data padding at boundaries of the image data. Additionally a method is proposed in which at least one of a transform and quantization process to be applied to de-multiplexed data is modified. One application is to combine left and right stereoscopic images, interleaved in a checkerboard manner.

Abstract: A system and method for compressed domain compression are provided for improving compression gains in an encoded image, such as a Joint Photographic Experts Group (JPEG)-encoded images, or encoded video, such as Motion Joint Photographic Experts Group (Motion JPEG)-encoded video, without fully decoding and re-encoding the compressed images or video.

Abstract: Computer implemented methods for compressing 3D hyperspectral image data having a plurality of spatial pixels associated with a hyperspectral image, and a number of spectral dimensions associated with each spatial pixel, include receiving, using a processor, the 3D hyperspectral image data, a set of basis vectors associated therewith, and either a maximum error amount or a maximum data size. The methods also include partitioning the 3D hyperspectral image data into a plurality of 2D images, each associated with one of the number of spectral dimensions, and an associated one of the set of basis vectors. The methods additionally include ranking the set of basis vectors if not already ranked. The methods may further include iteratively applying lossy compression to the 2D images, in an order determined by the ranking. Other embodiments and features are also disclosed.

Abstract: There is provided a method and a device for encoding a digital representation of an image into a single encoded image, comprising: generating at least two subsequent image data sequences each representing a portion of the digital representation of the image, inputting to an encoder (104) each of said subsequent image data sequences as if each image data sequence is an individual digital representations of an image, encoding each subsequent image data sequence into a variable length coded image, thereby generating at least two variable length coded images each representing a portion of the single encoded image, inserting a restart marker as terminating data of at least one of the variable length coded images, inserting an end of image marker as terminating data of one of the variable length coded images representing a final portion of the single encoded image, and associating the variable length coded images with each other by arranging them as subsequent parts of a single encoded data sequence representing th

Abstract: Each block of texture data elements is encoded as a block of texture data that includes: data indicating how to generate a set of data values to be used to generate data values for a set of the texture data elements that the block represents; data indicating a set of integer values to be used to generate the set of data values to be used to generate data values for a set of the texture data elements that the block represents; data indicating a set of index values indicating how to use the generated set of data values to generate data values for texture data elements of the set of texture data elements that the generated set of data values is to be used for; and data indicating the indexing scheme that has been used for the block.

Abstract: A scalable and high performance near-duplicate image search method utilizing short hashes improves performance over existing methods. By leveraging the shortness of the hashes, the search algorithm analyzes the reliability of each bit of a hash and performs content adaptive hash lookups by adaptively adjusting the “range” of each hash bit based on reliability. Matched features are post-processed to determine the final match results. The method can detect cropped, resized, print-scanned and re-encoded images and pieces from images among thousands of images.

Abstract: An exemplary embodiment of the invention relates to a method of using pattern vectors for image coding and decoding. The method comprises converting a block of image data into a set of transform coefficients, quantizing the transform coefficients such that a number of the coefficients become zero, constructing a single entity or bit vector indicating which coefficients are non-zero, coding the single entity or bit vector as an integer using an adaptive, semi-adaptive or non-adaptive arithmetic coder, coding the values of the coefficients in any fixed order, using an adaptive, semi-adaptive or non-adaptive arithmetic coder, or some other coder, and coding all coefficients except the zero coefficients. The system and method of decoding data relate to the corresponding hardware and process steps performed by the decoder when decoding a bitstream coded as described herein.

Abstract: Disclosed is a global motion detecting method which includes receiving a video sequence of input images, calculating local motion vectors, one for each image block of a current input image, grouping image blocks of the current input image into image block groups, calculating a group motion parameter of each of the image block groups based on local motion vectors of the image blocks in each respective image block group, and determining a global motion parameter of the currently input image according to the group motion parameters.

Abstract: A selective predictor utilizes a selective prediction method for coordinate data encoding. The selective predictor may calculate multiple prediction vectors for each vertex in the shape traversal order. Then, the selective predictor selects the vector that is closest to the vertex to be estimated. In determining the multiple prediction vectors, the selective predictor considers the position of several previous vertices in the traversal order. The selective predictor outputs a correction vector for the selected prediction vector and identification information indicting the selected prediction vector.

Abstract: A computer-implemented method and apparatus are disclosed for decoding an encoded data signal. In one embodiment, the method includes accessing, in a memory, a set of signal elements. The encoded data signal is received at a computing device. The signal includes signal fragments each having a projection value and an index value. The projection value has been calculated as a function of at least one signal element of the set of signal elements and at least a portion of the data signal. The index value associates its respective signal fragment with the at least one signal element used to calculate the projection value. The computing device determines amplitude values based on the projection values in the signal fragments. The decoded signal is determined using the amplitude values and the signal elements associated with the at least some of the signal fragments.

Abstract: Systems and methods are disclosed for providing sales or customer support using a telephone having a camera to capture at least a partial image of a product or service; wherein the at least partial image is used to select a database updated by one or more knowledgeable persons on the product or service.

Abstract: A difference of coordinate values stored adjacent to each other is compressed by means of a statistical coding system when reading out outline font data storing coordinate values necessary for drawing a contour of a character in order of drawing the contour in a clockwise or counterclockwise direction and also a category of a line connecting a pair of coordinates simultaneously, followed by compressing the coordinate values of the outline font data. A value of a result of subtracting “A?1” from a difference of coordinate values is determined to be a difference value of coordinates if the difference of coordinate value is equal to or greater than a certain value A, and a code expressing the difference value of “0” is added in front of the codes of difference values that are smaller than the value A in the case of a category of line connecting adjacent coordinates to each other being a straight line.

Abstract: Data representing animated hair in a computer generated imagery (CGI) scene may be compressed by treating hair data as arrays of parameters. Hair data parameters may include control vertices, hair color, hair radius, and the like. A principal component analysis (PCA) may be performed on the arrays of hair data. PCA may yield new basis vectors, varying in length, with the largest basis vector corresponding to a new dimension with the largest variance in hair data. The hair data may be quantized based on the varying lengths of new basis vectors. The number of bits allocated for quantizing each new dimension corresponding to each new basis vector may be determined based on the relative lengths of new basis vectors, with more bits allocated to dimensions corresponding to longer basis vectors. The quantized hair data may be bit-packed and then compressed using lossless entropy encoding.

Abstract: A method of noise filtering of a digital video sequence is provided that includes computing a motion image for a frame, wherein the motion image includes a motion value for each pixel in the frame, and wherein the motion values are computed as differences between pixel values in a luminance component of the frame and corresponding pixel values in a luminance component of a reference frame, applying a first spatial noise filter to the motion image to obtain a final motion image, computing a blending factor image for the frame, wherein the blending factor image includes a blending factor for each pixel in the frame, and wherein the blending factors are computed based on corresponding motion values in the final motion image, generating a filtered frame, wherein the blending factors are applied to corresponding pixel values in the reference frame and the frame, and outputting the filtered frame.

Abstract: In a learning process, first, images having different resolutions are obtained from a target region of the subject (S101). Further, the subject characteristic of the target region is obtained (S102). Then, the resolution conversion rules are learned from the images having different resolutions, and those are recorded to a storage device along with the subject characteristics (S103). When converting the resolutions, the resolution conversion rules learned for the corresponding subject characteristics are applied to each region of the original image so as to convert the resolutions of the original image.

Abstract: Various embodiments provide techniques and/or systems for reducing instances of computational complexity involved in video encoding.

Abstract: An exemplary embodiment of the invention relates to a method of using pattern vectors for image coding and decoding. The method comprises converting a block of image data into a set of transform coefficients, quantizing the transform coefficients such that a number of the coefficients become zero, constructing a single entity or bit vector indicating which coefficients are non-zero, coding the single entity or bit vector as an integer using an adaptive, semi-adaptive or non-adaptive arithmetic coder, coding the values of the coefficients in any fixed order, using an adaptive, semi-adaptive or non-adaptive arithmetic coder, or some other coder, and coding all coefficients except the zero coefficients. The system and method of decoding data relate to the corresponding hardware and process steps performed by the decoder when decoding a bitstream coded as described herein.

Abstract: The restoration of images by vector quantization utilizing visual patterns is disclosed. One disclosed embodiment comprises restoring detail in a transition region of an unrestored image, by first identifying the transition region and forming blurred visual pattern blocks. These blurred visual pattern blocks are compared to a pre-trained codebook, and a corresponding high-quality visual pattern blocks is obtained. The high-quality visual pattern block is then blended with the unrestored image to form a restored image.

Abstract: Quantization for oversampled signals with an error minimization searches based upon clusters of possible sampling vectors where the clusters have minimal correlation and thereby decrease reconstruction error as a function of oversampling (redundancy) ratio.

Abstract: A video processor is described, which is useful for implementing a quantization process, in compliance with the H.264 standard. The video processor includes an input, for receiving a block of image data. The image data is loaded into an internal register. In response to receiving a SIMD instruction, a quantizer, which incorporates the quantization lookup tables associated with the H.264 standard in its associated hardware, makes necessary high-level quantization decisions. In response to receiving another SIMD instruction, the quantizer uses those high-level quantization decisions to retrieve specific values from the quantization lookup tables.

Abstract: Local descriptors are extracted from an image. An image vector is generated having vector elements indicative of parameters of mixture model components of a mixture model representing the extracted local descriptors. The image vector is compressed using a vector quantization algorithm to generate a compressed image vector. Optionally, the compressing comprises splitting the image vector into a plurality of sub-vectors each including at least two vector elements, compressing each sub-vector independently using the vector quantization algorithm, and concatenating the compressed sub-vectors to generate the compressed image vector. Optionally, each sub-vector includes only vector elements indicative of parameters of a single mixture model component, and any sparse sub-vector whose vector elements are indicative of parameters of a mixture model component that does not represent any of the extracted local descriptors is not compressed.

Abstract: Image descriptor quantization technique embodiments are presented which quantize an image descriptor defined by a vector of number elements. This is generally accomplished by lowering the number of bits per number element to a prescribed degree. The resulting quantized image descriptor exhibits minimal loss of matching reliability while at the same time reducing the amount of storage space needed to store the descriptor in a database. Lowering the number of bits per number element also allows for increased matching speed.

Abstract: An image processing apparatus according to the present invention, comprises: a motion detection unit that detects a motion vector from an input image; a determination unit that determines whether an image is moving in each pixel in use of the detected motion vector, and determines whether a motion pixel, about which determination has been made that the image is moving therein, exists in a predetermined range from a still pixel about which determination has been made that the image is not moving therein; and a correction unit that performs correction processing to decrease at least one of high frequency components, contrast, and luminance for the still pixel about which determination has been made that a motion pixel exists in the predetermined range.

Abstract: Systems and methods for information compression. In a first method embodiment, a computer implemented method for compressing information includes finding a closest match entry of a dictionary to a data block, the dictionary stored in computer usable media. A logical exclusive OR combination of the closest match entry and the data block is performed to form a spatial difference vector identifying locations wherein the data block differs from the closest match entry. An information difference vector comprising differences between the data block and the closest match entry is formed. A compressed data block comprising the spatial difference vector, the information difference vector and a reference to the closest match entry is produced. The compressed data block is characterized as being smaller than the data block.