Abstract: Efficient inner product search is important for many data ranking services, such as recommendation and Information Retrieval. Efficient retrieval via inner product dramatically influences the performance of such data searching and retrieval systems. To resolve deficiencies of prior approaches, embodiments of a new index graph construction approach, referred to generally as Norm Adjusted Proximity Graph (NAPG), for approximate Maximum Inner Product Search (MIPS) are presented. With adjusting factors estimated on sampled data, NAPG embodiments select more meaningful data points to connect with when constructing a graph-based index for inner product search. Extensive experiments verify that the improved graph-based index pushes the state-of-the-art of inner product search forward greatly, in the trade-off between search efficiency and effectiveness.

Abstract: An audio signal encoding and decoding method using a learning model, a training method of the learning model, and an encoder and decoder that perform the method, are disclosed. The audio signal decoding method may include extracting a first residual signal and a first linear prediction coefficient by decoding a bitstream received from an encoder, generating a first audio signal from the first residual signal using the first linear prediction coefficient, generating a second linear prediction coefficients and a second residual signal from the first audio signal, obtaining a third linear prediction coefficient by inputting the second linear prediction coefficient into a trained learning model, and generating a second audio signal from the second residual signal using the third linear prediction coefficient.

Abstract: A video encoder including a first buffer containing a plurality of data values defining a macroblock of pixels of a video frame. The video encoder also includes a second buffer and an entropy encoder coupled to the first and second buffers and configured to encode a macroblock based on another macroblock. The entropy encoder identifies a subset of the data values from the first buffer defining a given macroblock and copies the identified subset to the second buffer, the subset of data values being just those data values used by the entropy encoder when subsequently encoding another macroblock.

Abstract: A first aspect is a method for coding a group of pictures (GOP) that includes frames of a video. The method includes encoding, at least some of the frames of the GOP, using a first encoding pass to obtain encoding statistics; obtaining, using the encoding statistics, respective temporal dependency likelihoods (TDLs) for the at least some of the frames of the GOP, where the respective TDLs indicate contributions that the at least some of the frames make in reducing prediction errors of the GOP; obtaining a reference frame based on the respective TDLs; and using the reference frame in encoding at least some of the frames of the GOP in a second encoding pass.

Abstract: A self-correcting secure computer system is provided. The computer system includes a read-only memory (ROM) device, a random access memory (RAM) device, and at least one processor in communication with the ROM device and the RAM device. The at least one processor is programmed to receive an activation signal; retrieve, from the ROM device, data to execute a first configuration including an encryption suite; execute, on the RAM device, the first configuration including the encryption suite; execute the encryption suite to generate a key; store the key at a first memory location; and delete volatile memory associated with the encryption suite.

Abstract: An electronic device, a method, and computer readable media for adapting a hierarchical codebook. The electronic device includes a memory for storing a hierarchical codebook and a processor operably connected to the memory, which is configured to determine usage probabilities for codewords in the hierarchical codebook. The codewords are represented by nodes assigned to one of a plurality of tiers of the hierarchical codebook based on a tier identifier. The processor is also configured to change a structure of the hierarchical codebook based on the usage probabilities so that nodes associated with codewords having relatively higher usage probabilities are located closer to a root node than nodes associated with codewords having relatively lower usage probabilities, and select a codeword from the hierarchical codebook for use in wireless communication. The codeword is selected by a codeword selection procedure performed on the changed structure of the hierarchical codebook.

Abstract: Provided is a method of reducing power consumption by a display device including an encoder for receiving a stream of data, and for compressing the data, a TX rate-buffer for receiving and storing the compressed data, a PHY for receiving the compressed data, a RX rate-buffer for receiving and storing the compressed data, and a decoder for receiving the compressed data, and for decompressing the compressed data to reconstruct original data, the method including placing the PHY a SLEEP state to reduce power consumption of the PHY when the TX rate-buffer transmits a last bit of the compressed data in the TX rate-buffer to the PHY, and placing the PHY in a TRANSMIT/ACTIVE state when a fullness of the TX rate-buffer reaches a reference threshold, or a last bit of compressed data corresponding to a last pixel of a line of pixels is placed in the TX rate-buffer.

Abstract: Motion compensation in a systems for transmission of images via channels of communication with a limited capacity by means of application of compression of the images. A technical result consists in an increase of the compression degree upon encoding such that a high degree of compression is provided without an increase in the computational power of the encoding device and without additional distortion upon decoding. The result is obtained in part by the usage of a more effective method of interpolation of restored subsamples, employing an adaptive and applicative set of samples, which are used to restore a quantized signal. Use of this set of samples facilitates an improvement in an accuracy of the interpolation.

Abstract: A system, apparatus, and method for decoding a video signal having a plurality of frames including a current frame. Decoding the video signal includes generating an alternate reference frame, detecting an error in the current frame and reporting the detected error, receiving a recovery frame in response to reporting the detected error, wherein the recovery frame is encoded using the alternate reference frame; and decoding the recovery frame using the alternate reference frame. The alternate reference frame occurs prior to the current frame in the video signal and is other than a frame immediately prior to the current frame.

Abstract: The present disclosure relates to analysing input data, prior to encoding, using one or more hierarchical algorithms. According to a first aspect, there is provided a method for producing output data using one or more input data and one or more hierarchical algorithms, comprising the steps of applying the hierarchical algorithm to the one or more input data; and producing output data to be used by an encoder; wherein one of the one or more input data is uncompressed; and wherein the output data is used to modify a decision making process associated with the encoder.

Abstract: The present disclosure discloses method and video generation system for generating video content based on user data. The video generation system receives user data sequentially from user, where each sequence of user data is converted into text data. One or more objects, relations, emotions, and actions from user data is identified by evaluating text data, a scene descriptor is generated for each sequence of user data, by associating one or more objects with one or more relations, emotions, and actions. The method comprises performing consistency check for scene descriptor of each sequence of user data, based on one or more previously stored scene descriptors, performing, one or more modifications to inconsistent scene descriptors, identified based on consistency check, generating, segments for each of scene descriptor and generating video content for by combining video segments associated with each of scene descriptor.

Abstract: Providing efficient lossless compression for small data blocks in processor-based systems is provided. In one aspect, a method comprises receiving a plurality of input words. Each mask of a plurality of masks is applied to each unassigned input word to generate a corresponding plurality of patterns. For each mask, if a most frequently occurring pattern exists among the plurality of patterns, the most frequently occurring pattern and an uncompressed data portion of each unassigned input word are stored in association with a prefix associated with the mask. The prefix is also assigned to each unassigned input word corresponding to the most frequently occurring pattern. A compressed output block is generated, comprising prefixes assigned to the plurality of input words, the most frequently occurring patterns associated with the assigned prefixes, and uncompressed data portions corresponding to one or more input words of the plurality of input words.

Abstract: Innovations in adaptive encoding for units of a video sequence can improve coding efficiency. For example, some of the innovations relate to encoding that includes adaptive switching of color spaces between units within a video sequence. Other innovations relate to encoding that includes adaptive switching of color sampling rates between units within a video sequence. Still other innovations relate to encoding that includes adaptive switching of bit depths between units within a video sequence.

Abstract: A method and apparatus for parallel processing of at least two bins relating to at least one of a video and an image. The method includes determining scan type of at least a portion of the at least one of video and an image, analyzing neighboring position of a bin, removing dependencies of context selection based on the scan type and position of location being encoded in a transform, and performing parallel processing of that least two bins.

Abstract: Methods and apparatuses for storing structured information are disclosed. A hash value computed over structured information determined for a host is compared to a hash value computed over a corresponding structured information stored in a remote database for the host where after an update of at least a part of the stored structured information can take place in response to determining a difference in the hash values and the stored structured information is kept in the database as it is in response to determining that the hash values are equal.

Abstract: Provided is a method and apparatus for segmenting airways and pulmonary lobes. An image processing apparatus obtains a first candidate region of an airway from a three-dimensional (3D) human body image by using a region growing method, obtains a second candidate region of the airway based on a directionality of a change in signal intensity of voxels belonging to a lung region segmented from the 3D human body image, segments an airway region by removing noise based on similarity of a directionality of a change in signal intensity of voxels belonging to a third candidate region acquired by combining together the first and second candidate regions.

Abstract: A switching method of video setting for a video conversion system includes generating a static metadata change flag when a static metadata change event is detected, repeating an output video after the static metadata change flag is generated, updating a video setting according to a static metadata corresponding to an input video when the output video is repeated, and converting the input video into the output video according to the video setting to display the output video.

Abstract: A method of processing an image is disclosed. The method comprises decomposing the image into a plurality of channels, each being characterized by a different depth-of-field, and accessing a computer readable medium storing an in-focus dictionary defined over a plurality of dictionary atoms, and an out-of-focus dictionary defined over a plurality of sets of dictionary atoms, each set corresponding to a different out-of-focus condition. The method also comprises computing one or more sparse representations of the decomposed image over the dictionaries.

Abstract: In a signal coding method, bits for coding allocated to different bands of a frequency domain signal obtained from an input signal are adjusted to improve the coding quality. The total available bits for coding are first allocated to the bands of the frequency domain signal according to a predetermined allocation rule. The numbers of bits allocated to the respective bands of the frequency domain signal are then adjusted when a highest frequency of the frequency domain signal to which bits are allocated is greater than a predetermined value. The frequency domain signal is coded according to the adjusted bit allocation for the bands of the frequency domain signal.

Abstract: Techniques for identifying and discriminating between different types of contacts to a multi-touch touch-screen device are described. Illustrative contact types include fingertips, thumbs, palms and cheeks. By way of example, thumb contacts may be distinguished from fingertip contacts using a patch eccentricity parameter. In addition, by non-linearly deemphasizing pixels in a touch-surface image, a reliable means of distinguishing between large objects (e.g., palms) from smaller objects (e.g., fingertips, thumbs and a stylus) is described.

Abstract: A decoder includes a memory storing a map of a space of encoded values. The map includes a plurality of cells partitioning the space, such that each cell encloses a cluster of encoded values and quantizes the cluster of encoded values to a quantized encoded value. Each cell is identified by a label selected from a finite alphabet, such that multiple cells in the map are identified by the same label, and the cells are labeled such that a pair of cells identified by the same label does not share a common boarder. The decoder also includes a receiver to receive from an encoder a label of a cell enclosing an encoded value on the map of the space and a processor to estimate the encoded value using side information to produce an estimation of the encoded value, to select a cell identified by the received label on the map of the multi-dimensional space that is the closest to the estimation of encoded value, and to determine the encoded value as the quantized encoded value of the selected cell.

Abstract: An image processing device includes an image sequence acquisition section that acquires an image sequence that includes a plurality of images, and a processing section that performs an image summarization process that deletes some of the plurality of images included in the image sequence acquired by the image sequence acquisition section to acquire a summary image sequence, the processing section selecting a reference image and a determination target image from the plurality of images, and determining whether or not the determination target image can be deleted based on the results of a process that utilizes deformation information about the reference image and the determination target image, and a process that utilizes a structural element that corresponds to an attention area.

Abstract: A method, apparatus, and computer readable medium for frame rate conversion based on object tracking. The method includes determining whether a block of a first frame of an input video stream corresponds to a detected object region or a non-object region. The method further includes selecting a motion estimation operation from a group of operations comprising an object tracking operation based at least on the determining of whether the block of the first frame corresponds to the detected object region or the non-object region. The method further includes determining a motion vector of the block by performing the selected motion estimation operation and determining an interpolated frame based at least on the determined motion vector and the first frame and a second frame of the input video stream.

Abstract: This disclosure describes techniques for coding transform coefficients for a block of video data. According to these techniques, a video coder (a video encoder or video decoder) determines whether a block of video data is a luma block or a chroma block. If the block of video data is a luma block, the video coder adaptively updates a VLC table index value based on a code number cn and value of a scaling factor. However, if the block of video data is a chroma block, the video coder adaptively updates the VLC table index value based on the code number cn and without using the scaling factor. The video coder uses the updated VLC table index value to select a VLC table of a plurality of VLC tables that are used to encode or decode the block of video data.

Abstract: A system to collect and store in a special data structure arranged for rapid searching massive amounts of data. Performance metric data is one example. The performance metric data is recorded in time-series measurements, converted into unicode, and arranged into a special data structure having one directory for every day which stores all the metric data collected that day. The data structure at the server where analysis is done has a subdirectory for every resource type. Each subdirectory contains text files of performance metric data values measured for attributes in a group of attributes to which said text file is dedicated. Each attribute has its own section and the performance metric data values are recorded in time series as unicode hex numbers as a comma delimited list. Analysis of the performance metric data is done using regular expressions.

Abstract: A method for parallel multi-dimensional encoding, the method may include receiving or generating a first version of a group of bits and a second version of the group of bits, wherein the first and second versions differ from each other by an arrangement of bits of the data unit; selecting a first set of bits of the first version and a second set of bits of the second version; encoding, in parallel, the first set of bits and the second set of bits; wherein the encoding of the second set of bits is responsive to the second set of bits and a first redundancy result of the encoding of the first set of bits; and wherein the encoding of the first set of bits is responsive to the first set of bits and to a second redundancy result of the encoding of the second set of bits.

Abstract: Methods and systems for detecting anomalies in transportation related video footage. In an offline training phase, receiving video footage of a traffic location can be received. Also, in an offline training phase, event encodings can be extracted from the video footage and collected or compiled into a training dictionary. One or more input video sequences captured at the traffic location or a similar traffic location can be received in an online detection phase. Then, an event encoding corresponding to the input video sequence can be extracted. The event encoding can be reconstructed with a low rank sparsity prior model applied with respect to the training dictionary. The reconstruction error between actual and reconstructed event encodings can then be computed in order to determine if an event thereof is anomalous by comparing the reconstruction error with a threshold.

Abstract: An image sensor can include pixels that are grouped into subsets of pixels, with each subset including three or more pixels. A method for asymmetrical high dynamic range imaging can include capturing an image of a subject scene using a single integration time for all of the pixels. In a subset of pixels, charge in N pixels is read out and summed together. N represents a number that is between two and one less than a total number of pixels in the subset. Un-summed charge is read out from one pixel in the subset. The un-summed charge and the summed charge are combined when producing a high dynamic range image.

Abstract: According to one exemplary embodiment, a method for image processing selects at least one image from a plurality of images, and stores the plurality of images into a buffer, until a storage space of the buffer reaches a criterion. When the storage space of the buffer reaches the criterion, a procedure of image processing is performed for a plurality of temporary images stored in the buffer, which includes constructing an image dictionary according to the at least an image selected, and performing an image recovery and reconstruction procedure for the plurality of temporary images in the buffer according to the image dictionary, thereby obtaining a plurality of reconstructed images.

Abstract: Approaches for deciding what individuals in a population of visual system “neurons” are looking for using sparse overcomplete feature dictionaries are provided. A sparse overcomplete feature dictionary may be learned for an image dataset and a local sparse representation of the image dataset may be built using the learned feature dictionary. A local maximum pooling operation may be applied on the local sparse representation to produce a translation-tolerant representation of the image dataset. An object may then be classified and/or clustered within the translation-tolerant representation of the image dataset using a supervised classification algorithm and/or an unsupervised clustering algorithm.

Abstract: A video signal fault detection system includes a front-end device and a back-end device. The back-end device includes: a quality evaluation unit generating a quality evaluation result associated with successive image frames captured by an image capturing device of the front-end device and transmitted through a transmission channel; a recognition unit generating, based on the quality evaluation result, a recognition output indicating a non-corruption condition or a corruption type of each transmitted image frame; and an inference unit inferring, based on a predetermined lookup table and a determination output generated based on the recognition output, whether each of the transmission channel and the image capturing device is normal or not.

Abstract: An automatic producing method for a predicted value generation procedure that predicts a value of an encoding target pixel by using a previously-decoded pixel. A parent population is generated by randomly producing predicted value generation procedures each of which is indicated by a tree structure, and a plurality of predicted value generation procedures are selected as parents from the parent population. One or more predicted value generation procedures are generated as children based on a predetermined tree structure developing method which subjects the selected predicted value generation procedures to a development where an existing predicted value generation function can be an end node of a tree.

Abstract: Methods and systems for automatically detecting multi-object anomalies at a traffic intersection utilizing a joint sparse reconstruction model. A first input video sequence at a first traffic location can be received and at least one normal event involving P moving objects (where P is greater than or equal to 1) can be identified in an offline training phase. The normal event in the first input video sequence can be assigned to at least one normal event class and a training dictionary suitable for joint sparse reconstruction can be built in the offline training phase. A second input video sequence captured at a second traffic location similar to the first traffic location can be received and at least one event involving P moving objects can be identified in an online detection phase.

Abstract: A vehicular display system having a camera for producing a video signal and a display device for displaying the video signal. The display device includes an integrity check to detect a defective video signal and alert the driver if a defective video signal has been detected.

Abstract: Apparatus and methods for scalable block pixel filtering are described. A block filtering instruction is issued to a processing element (PE) to initiate block pixel filtering hardware by causing at least one command and at least one parameter be sent to a command and control function associated with the PE. A block of pixels is fetched from a PE local memory to be stored in a register file of a hardware assist module. A sub-block of pixels is processed to generate sub-block parameters and the block of pixels is filtered in a horizontal/vertical edge filtering computation pipeline using the sub-block parameters.

Abstract: An image capture apparatus comprises a first shake detection unit configured to detect a shake of the image capture apparatus, a rotational shake calculation unit configured to calculate a rotational shake amount, a second shake detection unit configured to detect the shake of the image capture apparatus, a translational shake calculation unit configured to calculate a translational shake amount, a correction amount calculation unit configured to calculate, based on the rotational shake amount and the translational shake amount, a correction amount for correcting image blurring, and a correction unit configured to correct the image blurring, wherein the rotational shake calculation unit calculates a first translational motion component and perspective component, and wherein the correction amount calculation unit calculates a second translational motion component.

Abstract: An apparatus and method for encoding video data and an apparatus and method for decoding video data are provided. The encoding method includes: splitting a current picture into at least one maximum coding unit; determining a coded depth to output an encoding result by encoding at least one split region of the at least one maximum coding unit according to operating mode of coding tool, respectively, based on a relationship among a depth of at least one coding unit of the at least one maximum coding unit, a coding tool, and an operating mode, wherein the at least one split region is generated by hierarchically splitting the at least one maximum coding unit according to depths; and outputting a bitstream including encoded video data of the coded depth, information regarding a coded depth of at least one maximum coding unit, information regarding an encoding mode, and information regarding the relationship.

Abstract: A method and apparatus for determining an intra prediction mode of a coding unit. Candidate intra prediction modes of a chrominance component coding unit, which includes an intra prediction mode of a luminance component coding unit, are determined, and costs of the chrominance component coding unit according to the determined candidate intra prediction modes are compared to determine a minimum cost intra prediction mode to be the intra prediction mode of the chrominance component coding unit.

Abstract: An image encoding apparatus which encodes picture data is provided. The apparatus comprises an encoding unit configured to encode a picture to be encoded; a decoding unit configured to decode the encoded picture; an SN ratio calculation unit configured to calculate an SN ratio using the picture to be encoded and a decoding result of the decoding unit; a setting unit configured to set a target SN ratio serving as an index of the SN ratio; a bitrate control unit configured to control a bitrate of the picture to be encoded based on the target SN ratio; and a motion detection unit configured to detect motion information between the picture to be encoded and another picture, wherein the bitrate control unit controls the bitrate based on the motion information, and a difference between the SN ratio and the target SN ratio.

Abstract: A control method for an encoding apparatus which performs bit rate control includes an acquiring step of acquiring a first quantization parameter as a quantization parameter to be used for quantizing a rate control unit, a first calculating step of calculating an evaluation value from a preset rate and a rate of an actually encoded rate control unit, a first setting step of, if the evaluation value is within a predetermined range, setting the first quantization parameter acquired by the acquiring step as a quantization parameter to be used for quantizing a rate control unit, and a second setting step of, if the evaluation value is not within the predetermined range, setting a second quantization parameter on the basis of the calculated evaluation value as a quantization parameter to be used for quantizing a rate control unit.

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: In general, this disclosure describes techniques for encoding and decoding sequences of video frames using fragmentary reference pictures. The disclosure presents video encoding and decoding techniques for modified temporal compression based on fragmented references rather than complete reference pictures. In a typical sequence of video frames, only a portion (i.e., a tile) of each frame includes moving objects. Moreover, in each frame, the moving objects tend to be confined to specific areas that are common among each frame in the sequence of video frames. As described herein, such common areas of motion are identified. Pictures are then extracted from the identified areas of the video frames. Because these pictures may represent only portions of the frames, this disclosure refers to these pictures as “fragments.” It is then these fragments that are used as reference pictures for generating predicted frames during a motion compensation process, rather than the entire frame.

Abstract: An apparatus and method for encoding video data and an apparatus and method for decoding video data are provided. The encoding method includes: splitting a current picture into at least one maximum coding unit; determining a coded depth to output an encoding result by encoding at least one split region of the at least one maximum coding unit according to operating mode of coding tool, respectively, based on a relationship among a depth of at least one coding unit of the at least one maximum coding unit, a coding tool, and an operating mode, wherein the at least one split region is generated by hierarchically splitting the at least one maximum coding unit according to depths; and outputting a bitstream including encoded video data of the coded depth, information regarding a coded depth of at least one maximum coding unit, information regarding an encoding mode, and information regarding the relationship.

Abstract: An apparatus and method for encoding video data and an apparatus and method for decoding video data are provided. The encoding method includes: splitting a current picture into at least one maximum coding unit; determining a coded depth to output an encoding result by encoding at least one split region of the at least one maximum coding unit according to operating mode of coding tool, respectively, based on a relationship among a depth of at least one coding unit of the at least one maximum coding unit, a coding tool, and an operating mode, wherein the at least one split region is generated by hierarchically splitting the at least one maximum coding unit according to depths; and outputting a bitstream including encoded video data of the coded depth, information regarding a coded depth of at least one maximum coding unit, information regarding an encoding mode, and information regarding the relationship.

Abstract: An apparatus and method for encoding video data and an apparatus and method for decoding video data are provided. The encoding method includes: splitting a current picture into at least one maximum coding unit; determining a coded depth to output an encoding result by encoding at least one split region of the at least one maximum coding unit according to operating mode of coding tool, respectively, based on a relationship among a depth of at least one coding unit of the at least one maximum coding unit, a coding tool, and an operating mode, wherein the at least one split region is generated by hierarchically splitting the at least one maximum coding unit according to depths; and outputting a bitstream including encoded video data of the coded depth, information regarding a coded depth of at least one maximum coding unit, information regarding an encoding mode, and information regarding the relationship.

Abstract: An entropy coding module is provided for use in a video encoder that encodes a video input signal based on a plurality of macroblocks derived from the video input signal. The entropy coding module includes an entropy coder that generates entropy encoded data from discrete transformed coefficients for the plurality of macroblocks. A neighbor management module stores neighbor data for at least one macroblock of the plurality of macroblocks for retrieval by the entropy encoder, when operating on at least one neighboring macroblock of the plurality of macroblocks.

Abstract: The present invention provides an image encoding/decoding technique that is capable of achieving the higher compression efficiency. An image encoding method comprises: an intra prediction step which performs intra prediction on a block basis to generate a predicted image; a subtraction step which calculates the difference in prediction between the predicted image generated by the intra prediction step and an original image; a frequency conversion step which performs frequency conversion processing for the difference in prediction; a quantization step which subjects the output of the frequency conversion step to quantization processing; and a variable-length encoding step which subjects the output of the quantization step to variable-length encoding processing; wherein the intra prediction encoding step predicts a target pixel to be encoded by use of pixel values of two reference pixels between which the target pixel to be encoded is located.

Abstract: A method for rendering an MPEG sequence includes decoding a first picture in the MPEG sequence; reducing a data size of the decoded first picture by quantizing at least one component of the first picture, the quantized component selected from the luminance and chrominance components of the first picture; storing a reduced data size representation of the decoded first picture to a memory; reading a prediction block of the reduced data size representation of the decoded first picture; inverse quantizing the selected component of the prediction block of the reduced data size representation of the decoded first picture to restore the prediction block of the decoded first picture; and decoding a block of a second picture in the sequence according to the restored prediction block of the decoded first picture.

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: An image stabilization system includes: a region vector detection unit configured to detect a motion vector of each of a plurality of regions in an image; a reliability determination unit configured to determine reliability of the motion vector of each of the regions; a region indicator configured to indicate, to the region vector detection unit, a new detection target region instead of a region whose motion vector is determined to have low reliability; an image vector computing unit configured to compute a motion vector of the entire image using a motion vector determined to have high reliability; and a stabilization unit configured to move an entire image to be output according to the motion vector of the entire image to compensate for sway of the image.