Abstract: Embodiments of the present invention provide a computer-implemented method for adaptive residual gradient compression for training of a deep learning neural network (DNN). The method includes obtaining, by a first learner, a current gradient vector for a neural network layer of the DNN, in which the current gradient vector includes gradient weights of parameters of the neural network layer that are calculated from a mini-batch of training data. A current residue vector is generated that includes residual gradient weights for the mini-batch. A compressed current residue vector is generated based on dividing the residual gradient weights of the current residue vector into a plurality of bins of a uniform size and quantizing a subset of the residual gradient weights of one or more bins of the plurality of bins. The compressed current residue vector is then transmitted to a second learner of the plurality of learners or to a parameter server.

Abstract: Approaches for dynamic pre-filtering of digital video based on video complexity and output bit rate. An adaptive video preprocessor determines a current video complexity of the digital video and an output bit rate. Thereafter, the adaptive video preprocessor dynamically updates the strength of one or more preprocessing filters based on the current video complexity and the output bit rate for the digital video. The adaptive video preprocessor may update the strength of a preprocessing filter based, at least in part, upon selected values of a video quality preference category. A video quality preference category may be assigned natural language values which may each be translated into a particular strength value for at least one of the one or more preprocessing filters.

Abstract: Apparatus, system and method for tracking an image target in a system, wherein a system receives an image comprising a plurality of pixels. The received image is processed via a plurality of different recursive motion model kernels in parallel to provide a plurality of kernel outputs, wherein each of the motion model kernels may include a respective pixel mask. Per-pixel energy is estimated of at least some of the plurality of kernel outputs. Velocity of at least one of the image pixels may also be estimated by generating a directional energy vector for each motion model kernel. The per-pixel energy and velocity estimates are fused to produce a fused estimate representing at least some of the motion model kernels for the image.

Abstract: A method for organizing images from multiple image capture devices includes automatically determining a coarse offset between image capture times recorded in a first image capture device and image capture times recorded in a second image capture device. The coarse offset is determined by a computing a correlation between image counts of images captured by the first image capture device and images captured by the first image capture device. The method also includes adjusting the image capture times of images recorded in the second image capture device by the coarse offset to produce adjusted image capture times for images captured by the second image capture device.

Abstract: Apparatus and methods for managing the quality of experience of delivery of content to end-point or user devices. In one embodiment, metrics data pertaining to network conditions being experience by consumer devices attached to a managed network are monitored by a network entity. The monitored metrics are analyzed to determine if a consumer device is experiencing poor network conditions, such as network congestion. Consumer devices that are identified as experiencing poor network conditions are reported to a content delivery network (CDN) entity responsible for providing IP-based content. The CDN entity adjusts content delivery files used by the consumer devices to request particular IP-based content to guide the consumer device to select a version of the content suitable for delivery over consumer device's existing network condition.

Abstract: An apparatus having a circuit. The circuit may be configured to (i) calculate a plurality of complexity values while compressing a current picture in a video signal. Each complexity value generally characterizes how a corresponding one of a plurality of blocks in the current picture was compressed. The circuit may also be configured to (ii) adjust the complexity values below a first threshold to a default value and (iii) generate a region of interest by grouping the blocks having non-default values of the complexity values above a second threshold.

Abstract: Approaches for digital video encoding based, at least in part, on adaptive pre-filtering of the digital video. A user may select values for a plurality of video quality preference categories. An adaptive video preprocessor determines adaptive preprocessor information for the digital video, based, at least in part, on the current video complexity of the digital video and an output bit rate. The encoder encodes the digital video to produce encoded digital video using the adaptive preprocessor information. In this way, for example, the encoder may dynamically adjust how the digital video is encoded based on a selection, made by the user, regarding a stress bias video quality preference.

Abstract: Approaches for dynamic pre-filtering of digital video based on video complexity and output bit rate. An adaptive video preprocessor determines a current video complexity of the digital video and an output bit rate. Thereafter, the adaptive video preprocessor dynamically updates the strength of one or more preprocessing filters based on the current video complexity and the output bit rate for the digital video. The adaptive video preprocessor may update the strength of a preprocessing filter based, at least in part, upon selected values of a video quality preference category. A video quality preference category may be assigned natural language values which may each be translated into a particular strength value for at least one of the one or more preprocessing filters.

Abstract: In an embodiment a method of estimating a sparse signal from a measured signal using a transform matrix is disclosed. The method comprises determining a partial contribution to the sparse signal; determining a residual by subtracting the partial contribution from the measured signal; and iteratively determining further contributions to the sparse signal.

Abstract: A method for estimating the type of the GoP structure of a plurality of video frames in a video stream by estimating their frame types includes: capturing frame sizes in bytes of every video frame subsequent to an initial I-frame to obtain an array of frame sizes by exploiting features of a transport layer carrying the video stream; converting, after a number of frames, the array of frame sizes into an array of zeros and ones; matching the binarized array of frame sizes to a number of predefined short basic binary patterns, said predefined binary patterns depicting all GoP structures to be considered; converting the result of said matching to form a single score value; and determining the particular pattern of the number of predefined patterns of binaries having the best score value, according to a predefined metric.

Abstract: In a video coding apparatus, a plurality of coding units perform compression coding of a single source stream of video data. A control unit controls those coding units in such a way that all the coding units produce respective interframe coded pictures with identical playback times as a first-coded picture in a picture group (e.g., GOP) and produce identical local decoded pictures based on the respective intraframe coded pictures. A selection decision unit makes a selection, for each picture group, of one of the coded video data from the coding units on the basis of their coding results. According to this selection, a selection unit selectively outputs one of the coded video data. This output switching operation is performed at a point immediately before a forward interframe predictive coded picture that appears first of those in a picture group included in each of the coded video data outputs being switched.

Abstract: An encoder is operable to encode input data to generate corresponding encoded output data. The encoder includes data processing hardware. The encoder compresses content associated with blocks or packets, so that the encoded output data is smaller in size than the input data.

Abstract: An optical imaging system and associated methods for capturing images from an aircraft, such as a UAV. A camera unit on-board the aircraft is remotely controlled from an image control station. The image control station receives image data from the camera unit, and also delivers control signals for determining a viewing mode of the image.

Abstract: There are provided methods and apparatus for reducing coding artifacts for illumination compensation and/or color compensation in multi-view coded video. An apparatus includes an encoder for encoding at least one block in at least one picture for at least one view of multi-view video content. The encoder has a deblocking filter for performing adaptive deblocking filtering on the at least one block responsive to an indication of at least one of illumination compensation and color compensation being used for the at least one block.

Abstract: This invention makes it possible to reduce power consumption of an encoding device while more appropriately encoding a video image in the first frame. An image capturing apparatus includes a camera unit which photo-electrically converts object light and outputs a video signal, an encoding unit which encodes the video signal, a camera information acquisition unit which acquires information associated with the operation status of the camera unit, and a calculation unit which calculates, on the basis of the information acquired by the camera information acquisition unit, an initial parameter serving as an encoding parameter for an initial encoding operation of the encoding unit before the start of an encoding operation of the encoding unit.

Abstract: A video processing method is implemented by encoding and decoding devices. The video processing method includes: configuring the encoding device to decrease a resolution of a received at least one source image frame based on a received reference signal to obtain reduced image frame content with a reduced resolution, and to generate a relay image frame containing the reduced image frame content; configuring the encoding device to generate a header including at least one element containing information of the reduced image frame content, pack the relay image frame and the header into a video signal, and transmit the video signal to the decoding device; and configuring the decoding device to obtain a restored image frame from the reduced image frame content based on the element in the header. A data structure, and encoding and decoding devices are also disclosed.

Abstract: A computerized method for transforming a signal representative of image information, the signal being a sparse signal, includes obtaining the signal. The method further includes using the signal to learn a dictionary of basis functions that represent at least part of the signal. The method further includes using the learned dictionary of basis functions to transform the signal.

Abstract: A method and system for low-latency processing of intra-frame video pixel block prediction including: predicting a pixel block based on boundary pixels of left and upper neighbor blocks of said pixel block; subtracting said predicted pixel block from a source pixel block to generate a prediction error; forward transforming and quantizing said prediction error to generate a residual data; inverse transforming and quantizing said residual data; adding said predicted pixel block to said inverse transformed and quantized residual data to generate a reconstructed pixel block; pre-computing blocks of DC-coefficients used with luma and chroma intra prediction modes; pre-computing mode selection of a best prediction mode of said luma and chroma intra prediction modes; and outputting said residual data to be used in entropy or arithmetic coding, and a reconstructed data used for motion prediction.

Abstract: The disclosure is directed to techniques for region-of-interest (ROI) video processing based on low-complexity automatic ROI detection within video frames of video sequences. The low-complexity automatic ROI detection may be based on characteristics of video sensors within video communication devices. In other cases, the low-complexity automatic ROI detection may be based on motion information for a video frame and a different video frame of the video sequence. The disclosed techniques include a video processing technique capable of tuning and enhancing video sensor calibration, camera processing, ROI detection, and ROI video processing within a video communication device based on characteristics of a specific video sensor. The disclosed techniques also include a sensor-based ROI detection technique that uses video sensor statistics and camera processing side-information to improve ROI detection accuracy.

Abstract: A method for determining values of motion vectors includes receiving an irregular pattern of motion vectors for a target image, estimating an initial value for each of the motion vectors, using the motion vectors to generate a tap structure for an adaptive temporal prediction filter, and using the tap structure to re-estimate the value of each motion vector.

Abstract: A method of performing adaptive temporal prediction includes receiving a target image, wherein a position of an object in the target image is different from a position of the object in a reference image, using a boundary of the object to generate an irregular pattern of target image motion vectors, using the target image motion vectors to partition the target image into area of influence cells, wherein each area of influence cell contains a unique one of the target image motion vectors, and generating a prediction of the target image by applying an adaptive area of influence filter to the area of influence cells.

Abstract: A high-frequency integrating circuit 32 detects characteristics of horizontal high-frequency components and vertical high-frequency components of the image formed by a processing-target image signal. Based on the detection result, a CPU 61 obtains the number of bits of the after-compression-coding data of the image signal, and calculates the compression rate dependent upon the number of bits. The CPU 61 controls an image codec 36 so that the processing-target image signal is compression-coded through only one time of compression coding processing by use of the calculated compression rate. This configuration allows the image compression processing (compression coding) to be rapidly executed with high accuracy.

Abstract: An image coding apparatus executes intra-frame prediction with a predetermined intra-frame prediction mode for a certain block. A block selection unit judges that the current block to be coded is a prediction error calculation block or a non-prediction error calculation block. In the case where the current block is a prediction error calculation block, a mode error value calculation unit calculates the prediction error by the sum of absolute differences and for all of the intra-frame prediction modes, an inter-mode comparison unit compares the prediction errors, and as a result of the comparison, an intra-frame prediction mode having the minimum prediction errors is determined as the intra-frame prediction mode. In the case where the current block is a non-prediction error calculation block, the prediction error is not calculated. As an alternative, a prediction mode estimation unit determines the intra-frame prediction mode based on the neighboring prediction mode information stored in a storage unit.

Abstract: An image encoding method includes generating a predictive signal and encoding mode information according to each encoding mode from a macroblock signal corresponding to each macroblock, selecting a quantization code table corresponding to each macroblock, generating a predictive error signal for each encoding mode based on the macroblock signal and the predictive signal, subjecting the predictive error signal to orthogonal transformation, quantizing the orthogonal-transformed predictive error signal while changing a quantization parameter for every plural sub-pixel-blocks, using the quantization code table corresponding to the macroblock, encoding quantization transformation coefficient, calculating an encoding cost, selecting one encoding mode based on the encoding cost, selecting one quantization code table based on the encoding cost, and encoding information of an index indicating the selected quantization code table for every frame of the input image signal or every region of the frame.

Abstract: Low complexity (16 bit arithmetic) H.264 video compression replaces a single quantization table for all quantization parameters with multiple quantization tables and thereby equalizes quantization shifts and round-off additions; this eliminates the need for 32-bit accesses.

Abstract: An adaptive pre-processing technique is provided to reduce perceptually unimportant information in a frame (such as some of the high frequency information) so as to optimize compression efficiency and to maximize perceived video quality of a compressed video sequence for a given bit rate. This pre-processing includes a combination of performing pre-filtering of video data to remove high frequency components and to smooth out high frequency texture information within object boundaries, based at least in part on human visual system characteristics. The technique is adaptive in that feedback regarding the current state of the encoder is used to change the filtering strength or region of support for subsequent frames, as needed, in order to maintain the bit budget while still generating visually acceptable video quality.

Abstract: An image processing apparatus is adapted to a surveillance camera system, and a CPU included in the surveillance camera system divides a plurality of continuous screens of an image signal into a plurality of blocks by each screen, and detects a specific object, which is an object with movement, from a luminance change of a Y signal found by a movement detection circuit. In addition, the CPU specifies the block in which a movement of the object is detected, and sets the specified block as an area to be noticed by taking advantage of an ROI function of a JPEG 2000. Next, a JPEG 2000 CODEC is instructed to compress an image of a movement-detected block into a high quality image by an alarm compression rate, and compress the image of a movement-not-detected block by a normal compression rate having a higher compression rate than the alarm compression rate.

Abstract: An image processing apparatus is provided that can define the change amount of a quantizing parameter by suitably reflecting the complexity of image data in the case, when the quantizing parameter is increased a predetermined unit amount, the coarseness the image data of motion picture is quantized becomes r times. An activity computing circuit generates an activity Nactj serving as a complexity of the image data. A ?Q computing circuit defines a corresponding relationship between activity Nactj and change amount data ?Q such that the change amount data ?Q of quantizing parameter increases 1 as the activity Nactj becomes 1.12 times, to thereby acquire change amount data ?Q corresponding to the activity Nactj generated by the activity computing circuit.

Abstract: A process and apparatus for displaying and recording image data can be smoothly performed in characteristic methods by computing a quantization parameter for use in compression of image data, obtaining a difference in the amount of processes between the display of image data and the recording of image data, and managing memory of temporarily stored image data for regeneration of recorded image data.

Abstract: The image quality value of the code block currently processed is monitored, and when it falls below a threshold, encoding of the code block is terminated (in other words, encoding in the subsequent passes is omitted for the code block) to shift to encoding of the next code block.

Abstract: A method of measuring quality of service includes receiving, from a content server, a transmission of a first media stream and comparing that first media stream with a second media stream that corresponds to the first media stream prior to transmission thereof. This comparison provides a basis for determining a quality of service of the transmission.

Abstract: A quantization error compensation includes a dividing unit to divide an input current image signal into a first high-frequency signal and a low-frequency signal; a changing unit to replace bits of the first high-frequency signal with a predetermined signal to output a second high-frequency signal; an adding unit to add the low-frequency signal and the second high-frequency signal to create a composite signal; a quantization unit to cut n number of bits of the composite signal and to output the cut composite signal and the n number of bit signal; an equalizing unit to output a brightness equalizing value of the cut composite signal; a calculating unit to calculate a compensation value using the n number of the bit signal and a difference between brightness equalizing values with respect to a current and next brightness levels; and a compensation unit to add the compensation and the current brightness equalizing values.

Abstract: A method for quantization of a color histogram bin value of an image or video, and more particularly, a method for non-uniform quantization of a color histogram bin value of an image (or video) according to the frequency of color occurrence is provided. The methods effectively represent the characteristics of a color histogram of an image in comparison to conventional art, and improve the performance of image retrieval when an image (video) retrieval search is conducted.

Abstract: A method and system for compressing motion image information, which can compress data that can be subjected to predictive encoding. An image within a frame is divided into blocks before an inter-frame compression procedure begins, and each block is approximated with a single plane represented by at least three components for pixels within each block. Pixels between the original image and the image expanded after compressed can be compared, and when a pixel that causes greater difference than a given parameter to occur exists, intra-frame compression is performed using a smaller block size. Furthermore, when the respective I blocks, which are spatially divided, are dispersed between each frame along the temporal axis, no I block is inserted into any block within the frame that has been updated due to difference between frames being greater than parameter P during a designated period of time.

Abstract: An image signal processing apparatus for quantizing an inputted moving image signal according to a quantization step so that a code amount of one frame becomes a target code amount, and for variable-length-coding the quantized moving image signal, in which a minimum value of the quantization step is determined for each frame, and the quantization step is determined so as not to be less than the minimum value.

Abstract: The present invention relates to a method for quantization of a color histogram bin value of an image or video, and more particularly, to a method for non-uniform quantization of a color histogram bin value of an image (or video) according to the frequency of color occurrence. The present invention has effects of effectively representing the characteristics of the color histogram of the image better as compared to the conventional art, and improving the performance of the image retrieval when the image (video) retrieval search is conducted.

Abstract: A transformation unit 102 transforms an entered image to multi-resolution space. A quantizer 105 performs vector quantization on a local pattern of an image of a multi-resolution representation. A perspective-order calculating unit 107 extracts a plurality of code words, positions corresponding thereto and/or angle of rotation and/or scale, and the perspective-order relationship of a plurality of these representative vectors, from the quantized image. An algebraic encoder 108 encodes the input image based upon the extracted information. As a result, there are provided an image processing apparatus and method for asymmetric encoding without motion in three dimensions and extraction of a three-dimensional structure. Further, a transformation unit 1103 transforms an entered image to vector field, and a singularity detector 1104 detects a singularity in the transformed image.

Abstract: A method and apparatus for adaptive pixel estimation under high error rate conditions are disclosed. Classes are created, feasible classes are determined, and an optimal filter is applied to perform the pixel estimation.

Abstract: A video encoding method and a computer-readable record media for performing the method are disclosed, in which in carrying out the video encoding, an intra updating technique is adopted by sorting out the blocks requiring an intra updating by utilizing the bit stream size information for each block, thereby maximizing the intra updating effect. The video encoding method includes the following steps. That is, at a first step, an encoding method for each frame of incoming images is selected, and the images are divided into blocks of a certain size. At a second step, either an intra coded block encoding or an inter coded block encoding is carried out depending on whether an encoding of a current frame and an inter coded encoding are needed or not, and whether an intra coded block is relevant or not. Further, blocks requiring an intra updating are sorted out by utilizing a bit stream size information for each block if the inter coded encoding is needed.

Abstract: An apparatus and method for dynamically varying the frame rate of an image sequence is disclosed. In one embodiment, the image sequence is encoded and stored at different frame rates (e.g., 30, 25, 20 fps and so on). Alternatively, only the motion information, e.g., motion vectors, for the other frame rates are stored.

Abstract: A method for adaptive quantization of video frames based on bit rate prediction that includes increasing quantization in sectors of a video frame where coding artifacts would be less noticeable to the human visual system and decreasing quantization in sectors where coding artifacts would be more noticeable to the human visual system. In one embodiment the method reverts to uniform quantization for video frames in which adaptive quantization would require extra bits.

Abstract: The invention relates to a method of coding motion information associated to an image sequence, comprising the steps of subdividing each image into blocks and applying to each block a block-matching algorithm for defining a shifted block as the prediction of the current block, the motion vector between said shifted and current blocks being the predicted vector associated to said current block and all the motion vectors similarly predicted for a whole current image constituting a motion vector field associated to said current image. For each current image, the motion information constituted by said associated motion vector field is finally coded. According to the invention, the motion vector C to be coded is approximated by a spatio-temporal predictor P defined by a relation of the type: P=&agr;·S+&bgr;·T, where S and T are spatial and temporal predictors respectively, and (&agr;, &bgr;) are weighting coefficients respectively associated to said spatial and temporal predictors.

Abstract: A method for digital image compression of a raster image is disclosed which uses different compression methods for selected parts of the image and which dynamically adjusts compression and segmentation parameters to control tradeoff of image quality and compression. The image is encoded into a single data stream for efficient handling by disk, memory and I/O systems. This system provides a stable feedback loop that manages to a fine granularity (typically 8×8 pixel blocks) the compression of image data so as to be compliant with a plurality of compression constraints; the management of an incremental bias toward more aggressive compression that selectively applies the more aggressive compression methods to blocks that are the most tolerant with respect to visible artifacts; and the adjustable segmentation and compression parameters used to control the image compression.

Abstract: An image coding apparatus including a context generator configured to select a reference pixel that has been coded in image data to generate a context formed of the selected reference pixel, an entropy encoder configured to code the image data based on the context to generate coded data, an evaluation value calculator configured to calculating a compression ratio based on the image data and the coded data to output the compression ratio, and an adaptive evolution processor configured to perform adaptive evolution based on the compression ratio to find optimal reference pixels for the image data, the optimal reference pixels being supplied to the context generator for optimizing the context.

Abstract: A data organization and access scheme for permitting easier location of a reference macroblock during motion compensation and for providing a video output system with output data. The frames are segmented into coding blocks comprising a small number of horizontally adjunct blocks compressed to a fixed length. Within each coding block, only two quantization strategies are employed so that the start and end position of any coding block is easily ascertained as well as the location of any individual block within the coding block.

Abstract: A compression system and process employs a group of quantizers (or set of predefined quantized values) and involves the selection of the quantizers for each video frame or frame portion. For each frame portion, a selection of the most appropriate quantizer is made. The selection of which quantizer from the selection group is most appropriate for coding of a video frame or frame portion is based on a formula which takes account both the distortion (accuracy) and bit rate characteristics of each quantizer. The quantizer that exhibits the best combined distortion and bit rate characteristics is selected for coding the frame or frame portion. A similar formula, based on both distortion and bit rate characteristics, is used to select the particular quantization value within the quantizer set for each video signal value being coded.

Abstract: The variable bitrate coding method according to the invention includes an iterative process including a first coding pass and a second prediction pass. The coding pass allows for a picture coding of a sequence with a constant quantization stepsize (and quality) and the prediction pass allows for a matching of the stepsize to the wanted target bitrate. After some iterations, a last step allows for a finer adjustment of said stepsize with respect to said target bitrate.

Abstract: An apparatus for transcoding a coded moving picture sequence according to the present invention comprises a rate controller for adapting the scaling factor to perform the rate control over the coded moving picture sequence signal having a desired target bit rate.

Abstract: An apparatus and method for improving the delivery of a digital multimedia stream over a lossy packet network. The method consists in creating data packets of equivalent perceptual relevance to the end-user and as of equal length as possible. Therefore a packet loss induces the same perceptual degradation independently of its location in the multimedia stream.

Abstract: A technique for reducing computational and storage requirements of cascaded polyphase DFT-filter bank for receiving and transmitting telecommunications is disclosed. The cascaded polyphase DFT-filter bank is designed by specifically selecting a range of radio spectrum for reducing computational and storage requirements of the cascaded polyphase DFT-filter bank operation.