Abstract: A video processing method includes obtaining motion information of a neighboring block of a current image block, dividing the current image block into a plurality of sub-blocks in response to the neighboring block satisfying a preset condition, determining, in a time-domain reference image of the current image block, related blocks of the plurality of sub-blocks according to a motion vector of the neighboring block, and performing prediction on the current image block according to motion vectors of the related blocks.

Abstract: A method of compressing a frame in an image compression and storage system includes mapping an original sample to a mapped sample based on a bit depth of the original sample and a maximum allowed error, determining a residue of the mapped sample based on a mapped previous reconstructed sample, applying a modulo addition to the residue to generate a biased residue, quantizing the biased residue based on the maximum allowed error to generate a quantized biased residue, and encoding a value corresponding to the quantized biased residue to generate an encoded value.

Abstract: The present disclosure provides an encoding apparatus and a decoding apparatus, as well as an encoding method and a decoding method. In particular, the present disclosure relates to block partitioning and signaling the partitioning parameters in a bitstream. An override flag in a slice header indicates whether or not a block is to be partitioned according to a first partition constraint information. The override flag is included in the bitstream and the block is partitioned accordingly.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer readable media for generating a modified digital image from extracted spatial and global codes. For example, the disclosed systems can utilize a global and spatial autoencoder to extract spatial codes and global codes from digital images. The disclosed systems can further utilize the global and spatial autoencoder to generate a modified digital image by combining extracted spatial and global codes in various ways for various applications such as style swapping, style blending, and attribute editing.

Abstract: The present disclosure provides a method for constructing a motion information candidate list, a method and an apparatus for triangle prediction decoding. The method includes: determining a checking order for each of spatial candidate blocks of a current image block; wherein the checking order for the spatial candidate block on top of the current image block is prior to the checking order for the spatial candidate block on the left of the current image block; determining spatial candidate motion information based on the checking order; and constructing a motion information candidate list corresponding to the current image block based on the spatial candidate motion information. The method can reduce coding consumption, and improve coding performance.

Abstract: A video decoding method according to an embodiment of the present invention may include determining a type of a filter to be applied to a first-layer picture which a second-layer picture as a decoding target refers to; determining a filtering target of the first-layer picture to which the filter is applied; filtering the filtering target based on the type of the filter; and adding the filtered first-layer picture to a second-layer reference picture list. Accordingly, the video decoding method and an apparatus using the same may reduce a prediction error in an upper layer and enhance encoding efficiency.

Abstract: An image encoding/decoding method of the present invention constructs a merge candidate list of a current block, derives motion information of the current block on the basis of the merge candidate list and a merge candidate index, and performs inter prediction on the current block on the basis of the derived motion information, wherein the merge candidate list can improve encoding/decoding efficiency by adaptively determining a plurality of merge candidates on the basis of the position or size of a merge estimation region (MER) to which the current block belongs.

Abstract: A method for video decoding is disclosed. A directional intra prediction mode is decoded for a block of a picture, the directional intra prediction mode having a direction. Based on the directional intra prediction mode, a first predictor and a second predictor for a sample in the current block are accessed, the first and second predictors being on a line at least approximating the direction. A sample value of the sample is predicted by interpolation using the first and second predictors, responsive to a difference between the second and first predictors, wherein the difference is scaled by at least a ratio that is based on a location of the sample in the block, and wherein a denominator of the ratio is a power of two regardless of the location of the sample. The sample of the block is reconstructed based on the predicted sample value.

Abstract: Method and apparatus of Inter prediction for video coding are disclosed. When a sub-block motion compensation coding tool is selected for the current block, the method generates sub-block MVs (motion vectors) associated with multiple sub-blocks, which are included or contained in the current block, according to the sub-block motion compensation coding tool, constrains the sub-block MVs within a range to form constrained sub-block MVs, and applies motion compensation to the current block using the constrained sub-block MVs or applies motion compensation to the current block using one sub-block MV within the range around the primary MV in a second list if a corresponding sub-block MV in a first list is outside the range. In another method, motion compensation is applied to the current block only using reference pixels of reference sub-blocks within a primary reference block.

Abstract: Devices, systems and methods for digital video coding, which includes using default motion candidates for video coding, are described. An exemplary method for video processing includes determining, for a conversion between a current block of a video and a bitstream representation of the video, one or more motion candidates for a geometry partition mode using one or more pairwise average candidates that are obtained by averaging a pair of candidates in a motion candidate list for the current block. The geometry partition mode includes partitioning the current block into multiple partitions. The method also includes performing the conversion based on the determining.

Abstract: A method of adjusting a bitrate of an image based includes: identifying magnitudes of one or more frequency components in the image; determining a frequency filter based on a difference between a current bitrate and a target bitrate of the image such that the greater is the difference between the current bitrate and the target bitrate, magnitudes of the more high frequency components are reduced by the frequency filter; and applying the determined frequency filter to the magnitudes of the one or more frequency components.

Abstract: Provided is a video decoding method including determining a displacement vector per unit time of pixels of a current block in a horizontal direction or a vertical direction, the pixels including a pixel adjacent to an inside of a boundary of the current block, by using values about reference pixels included in a first reference block and a second reference block, without using a stored value about a pixel located outside boundaries of the first reference block and the second reference block; and obtaining a prediction block of the current block by performing block-unit motion compensation and pixel group unit motion compensation on the current block by using a gradient value in the horizontal direction or the vertical direction of a first corresponding reference pixel in the first reference block which corresponds to a current pixel included in a current pixel group in the current block, a gradient value in the horizontal direction or the vertical direction of a second corresponding reference pixel in the sec

Abstract: A merge candidate list is generated, a merge candidate is selected from the merge candidate list as a merge candidate, a bitstream is decoded to derive a motion vector difference, and a corrected merge candidate is derived by adding the motion vector difference to a motion vector of the selected merge candidate for a first prediction without scaling and subtracting the motion vector difference from a motion vector of the selected merge candidate for a second prediction without scaling.

Abstract: Provided is a video decoding method including: obtaining a first motion vector indicating a first reference block of a current block in a first reference picture and a second motion vector indicating a second reference block of the current block in a second reference picture; obtaining a parameter related to pixel group unit motion compensation of the current block, based on at least one of information of the parameter related to the pixel group unit motion compensation and a parameter related to an image including the current picture; generating a prediction block by performing, with respect to the current block, block unit motion compensation based on the first motion vector and the second motion vector and performing the pixel group unit motion compensation based on the parameter related to the pixel group unit motion compensation; and reconstructing the current block. Here, a pixel group may include at least one pixel.

Abstract: A decoder includes circuitry configured to receive a bitstream; construct, for a current block, a motion vector candidate list including adding a single global motion vector candidate to the motion vector candidate list, the single global motion vector candidate selected based on a global motion model utilized by the current block; and reconstruct pixel data of the current block and using the motion vector candidate list. Related apparatus, systems, techniques and articles are also described.

Abstract: A method of deriving a motion vector difference for coding video data, the method comprises determining whether inter-layer prediction is utilized by determining whether both a reference picture in a first reference picture list is not marked as used for long-term reference and a reference picture in a second reference picture list is not marked as used for long-term reference; based on whether the inter-layer prediction is utilized, setting a scaling factor to one of: a value of 1 in a case where the inter-layer prediction is utilized; or a value in the range of ?4096 to 4095 in a case where the inter-layer prediction is not utilized; and deriving the motion vector difference based on the scaling factor.

Abstract: A system and method for stitching separately encoded MPEG video fragments, each representing a different rectangular area of the screen together into one single full-screen MPEG encoded video fragment.

Abstract: Systems and methods are described for mobile and augmented reality-based depth and thermal fusion scan imaging. Some embodiments of the present technology use sophisticated techniques to fuse information from both thermal and depth imaging channels together to achieve synergistic effects for object recognition and personal identification. Hence, the techniques used in various embodiments provide a much better solution for, say, first responders, disaster relief agents, search and rescue, and law enforcement officials to gather more detailed forensic data. Some embodiments provide a series of unique features including small size, wearable devices, and ability to feed fused depth and thermal streams into AR glasses. In addition, some embodiments use a two-layer architecture for performing device local fusion and cloud-based platform for integration of data from multiple devices and cross-scene analysis and reconstruction.

Abstract: An image encoding method and an image decoding method are provided. The image decoding method includes deriving a temporal merge candidate from a co-located block of a current block, generating a merge candidate list of the current block based on the derived temporal merge candidate, and generating a prediction block of the current block based on the generated merge candidate list. The deriving a temporal merge candidate includes scaling a motion vector derived from the co-located block based on a POC difference value between the current block and a reference picture of the current block and a POC difference value between the co-located block and a reference picture of the co-located block, and modifying the scaled motion vector based on motion vector scaling information between a neighboring block of the current block and a co-located block of the neighboring block.

Abstract: A method, system, medium, and article provide neural network loop filtering for video coding with multiple alternative neural networks.

Abstract: The processor obtains a first pixel value and a second pixel value of the display. The processor determines a desired pixel value range that exceeds the second pixel value of the display. The processor obtains a threshold between the first pixel value of the display and the second pixel value of the display. The processor obtains a function mapping the desired pixel value range to a range between the threshold and the second pixel value. The processor applies the first function to an input image prior to displaying the input image on the display. The display presents the image. Upon recording the presented image, the processor determines a region within the recorded image having a pixel value between the threshold and the second pixel value. The processor increases dynamic range of the recorded image by applying an inverse of the function to the pixel value of the region.

Abstract: A method of decoding a coded video bitstream includes obtaining a sequence parameter set (SPS)-level syntax element from the bitstream, wherein that the SPS-level syntax element equals to a preset value specifies that no video parameter set (VPS) is referred to by a SPS, and the SPS-level syntax element greater than the preset value specifies that the SPS refers to a VPS, obtaining, as the SPS-level syntax element is greater than the preset value, an inter-layer enabled syntax element specifying whether one or more inter-layer reference pictures (ILRPs) are enabled to be used for the inter prediction of one or more coded pictures, and predicting one or more coded pictures based on the value of the inter-layer enabled syntax element.

Abstract: A video decoder can be configured to determine that a block of the video data is formatted in accordance with a 4:4:4 video coding format; determine that the block of the video data is encoded in an intra prediction mode; determine that a smallest chroma intra prediction unit (SCIPU) is disabled for the block in response to determining that the block has the 4:4:4 video coding format; decode the block of the video data based on the determination that the SCIPU is disabled; and output decoded video data comprising a decoded version of the block.

Abstract: A method and an apparatus are provided for obtaining a motion vector, and disclose application of the method and the apparatus for obtaining a motion vector in a video coding method and apparatus. The method includes: obtaining a first motion vector from a first motion vector set corresponding to a to-be-processed picture block; obtaining a second motion vector through calculation based on the first motion vector, where there is a preset vector difference between the first motion vector and the second motion vector, and the first motion vector and the second motion vector are different; adding the second motion vector to the first motion vector set to constitute a second motion vector set; and obtaining a target motion vector from the second motion vector set.

Abstract: A method is provided for encoding a digital video to improve perceptual quality. The method includes receiving a digital video at a video encoder, providing a perceptual quantizer function defined by P ? Q ? ( L ) = ( c 1 + c 2 ? L m 1 1 + c 3 ? L m 1 ) m 2 , wherein L is a luminance value, c1, c2, c3, and m1 are parameters with fixed values, and m2 is a parameter with a variable value, adapting the perceptual quantizer function by adjusting the value of the m2 parameter based on different luminance value ranges found within a coding level of the digital video, encoding the digital video into a bitstream using, in part, the perceptual quantizer function, transmitting the bitstream to a decoder, and transmitting the value of the m2 parameter to the decoder for each luminance value range in the coding level.

Abstract: Techniques are described to unify the motion vectors of sub-blocks of an affine block and the motion vector information that is stored and used as predictors for determining motion vectors for subsequent blocks. A video coder may determine that the motion vector for a sub-block is the same as the motion vector information that is used to determine motion vectors for subsequent blocks.

Abstract: A method of decoding a video sequence coded with subpictures can include receiving a first syntax element corresponding to a coding tree unit (CTU) size in a bitstream at a decoder, receiving a second syntax element corresponding to whether a picture being partitioned into subpictures, in response to the second syntax element indicating the picture is partitioned into the subpictures, receiving third syntax elements indicating a position and a size of each of the subpictures in the picture based on a unit of the CTU size, and reconstructing one or more of the subpictures based on the position and the size of each of the respective subpictures in the picture.

Abstract: Decoding a current block of a current frame includes selecting a first reference frame for forward inter prediction of the current frame; selecting a second reference frame for backward inter prediction of the current frame; generating an optical flow reference frame portion for inter prediction of the current block by performing an optical flow estimation using the first reference frame and the second reference frame, where the optical flow estimation produces a respective motion field for pixels of the current block; and performing a prediction process for the current block using the optical flow reference frame portion by: using a motion vector used to encode the current block to identify a reference block; adjusting boundaries of the reference block using a subpixel interpolation filter length; and identifying blocks encompassing pixels within the adjusted boundaries of the reference block.

Abstract: A video processing method includes performing or skipping following processing based on whether a size of a coding unit is not smaller or is smaller than 8×8. The processing includes dividing the coding unit into sub-blocks each having a fixed size of 8×8, scanning a left neighboring block and determining a reference motion vector of the coding unit, determining a related reference block of a sub-block in the co-located reference image according to the reference motion vector, determining a scaling factor of a motion vector of the related reference block, scaling the motion vector of the related reference block using the scaling factor, determining motion information of the sub-block according to the scaled motion vector, and performing prediction for the coding unit according to the motion information.

Abstract: A stream decompression circuit is disclosed. The stream decompression circuit includes a coding length first-in-first-out (FIFO) and a calculation circuit. The coding length FIFO is coupled to a variable length coding (VLC) circuit and used to store a coding length that the VLC circuit codes sub-streams and output a specific number of bits when the coding length accumulates over the specific number of bits. The calculation circuit is coupled between the coding length FIFO and a multiplexer circuit and used to calculate a number of bits required for decompression and output an output multiplex control signal to the multiplexer circuit to control the multiplexer circuit to output the sub-streams according to a specific order.

Abstract: A video processing method includes performing or skipping following processing based on whether a size of a coding unit is not smaller or is smaller than 8×8. The processing includes dividing the coding unit into sub-blocks each having a fixed size of 8×8, scanning a left neighboring block and determining a reference motion vector of the coding unit, determining a related reference block of a sub-block in the co-located reference image according to the reference motion vector, determining a scaling factor of a motion vector of the related reference block, scaling the motion vector of the related reference block using the scaling factor, determining motion information of the sub-block according to the scaled motion vector, and performing prediction for the coding unit according to the motion information.

Abstract: A moving picture coding apparatus includes: an inter prediction control unit which determines to code a motion vector using, among candidate predicted motion vectors, a candidate predicted motion vector having the least error with relative to a motion vector derived by motion estimation; a picture type determination unit which generates picture type information; a temporal direction vector calculation unit which derives a candidate predicted motion vector in temporal direct; and a co-located reference direction determination unit which generates, for each picture, a co-located reference direction flag.

Abstract: Data that is predicted across pictures in a video sequence is managed by separating the data into multiple data types. Instead of keeping all data associated with a decoded picture, such as picture sample values and motion vector data, data associated with a decoded picture is split by data type to enable storing only a subset of all data associated with a decoded picture.

Abstract: Systems, apparatuses, and methods for reducing latency when consuming an encoded video bitstream in real-time are disclosed. A video encoder encodes a video bitstream and writes chunks of the encoded bitstream to a bitstream buffer. Prior to the encoder completing the encoding of an entire frame, or an entire slice of a frame, a consumer module consumes encoded chunks of the bitstream. In one implementation, to enable pipelining of the consumption with the encoding, the encoder updates a buffer write pointer with an indication of the amount of data that has been written to the bitstream buffer. The consumer module retrieves encoded data from the bitstream buffer up to the location indicated by the buffer write pointer. In this way, the consumer module is able to access and consume encoded video data prior to the encoder finishing encoding an entire frame or an entire slice of the frame.

Abstract: A method of coding 4:2:2 or 4:4:4 video data comprises predicting luminance and/or chrominance samples of an image from other respective reference samples derived from the same image according to a prediction mode associated with a sample to be predicted, the prediction mode being selected for each of a plurality of blocks of samples, from a set of two or more candidate prediction modes; detecting differences between the samples and the respective predicted samples; selecting a frequency-separation transform from two or more candidate frequency separation transforms according to the prediction mode associated with a current block of samples using a mapping between transform and prediction mode, the mapping between different, as between chrominance and luminance samples, for at least the 4:4:4 format; and encoding the detected differences by frequency-separating the differences, using the selected frequency-separation transform.

Abstract: After initial intra prediction based on projection along an intra prediction direction, there may exist discontinuities between the predicted block and neighboring blocks. To smooth the block boundary, boundary filtering may be used with intra prediction. In particular, different methods may be used for boundary filtering. The particular filtering method to be used may be selected adaptively based on the intra prediction mode. When gradient-based filtering is selected, the gradient at a reference sample, that is at the same row or column as the current sample, in a reference layer is calculated as the difference between the reference sample and the predicted value for the reference sample using the intra prediction mode. Additional gradient at another reference sample from another reference layer may also be used. Then the initially predicted value for the sample may be adjusted based on the gradient(s) and a decay function.

Abstract: A reference picture marking process and a reference picture list management process is handled in a unified reference picture marking and reference picture list management process. A new idle reference picture list may be used for handling reference pictures that are not used for reference in the current picture. Differential coding of picture order count may be used to increase coding efficiency. The reference picture management syntax structure may be sent in the picture parameter set for improved coding efficiency e.g. in regular GOP (group of pictures) arrangements.

Abstract: A method for video decoding is disclosed. A directional intra prediction mode is decoded for a block of a picture, the directional intra prediction mode having a direction. Based on the directional intra prediction mode, a first predictor and a second predictor for a sample in the current block are accessed, the first and second predictors being on a line at least approximating the direction. A sample value of the sample is predicted by interpolation using the first and second predictors, responsive to a difference between the second and first predictors, wherein the difference is scaled by at least a ratio that is based on a location of the sample in the block, and wherein a denominator of the ratio is a power of two regardless of the location of the sample. The sample of the block is reconstructed based on the predicted sample value.

Abstract: An approach is provided for generating a reverse sequence or real-time streamed images for triangulation. The approach includes receiving a real-time stream of images captured by a sensor of a vehicle during a drive; extracting a sequence of two or more images from the real-time stream; reversing the sequence of the two or more images; and providing the reversed sequence of the two or more images for feature triangulation.

Abstract: A method comprising: encoding an input picture into a coded constituent picture (1200); reconstructing, as a part of said encoding, a decoded constituent picture corresponding to the coded constituent picture (1202); encoding a spatial region into a coded tile (1204), the encoding comprising: determining a horizontal offset and a vertical offset indicative of a region-wise anchor position of the spatial region within the decoded constituent picture (1204a); encoding the horizontal offset and the vertical offset (1204b); determining that a prediction unit at position of a first horizontal coordinate and a first vertical coordinate of the coded tile is predicted relative to the region-wise anchor position, wherein the first horizontal coordinate and the first vertical coordinate are horizontal and vertical coordinates, respectively, within the spatial region (1204c); indicating that the prediction unit is predicted relative to a prediction-unit anchor position that is relative to the horizontal and vertical off

Abstract: A video decoder configured to generate a first merge candidate list for a first block; determine that the first block is coded in a merge mode with motion vector differences; in response to determining that a maximum number of entries for the first merge candidate list is equal to 1, infer a value of a first instance of a flag to be equal to a first value, wherein the first value for the flag indicates that the first block is to be decoded using a first entry in the first merge candidate list; receive first motion vector difference information; determine first motion information for predicting the first block based on candidate motion information included in the first entry of the first merge candidate list and the first motion vector difference information; and decode the first block using the first motion information.

Abstract: A method of decoding video data includes predicting luma samples for a block of the video data using matrix intra prediction (MIP), where using MIP comprises down-sampling a set of luma neighboring samples to generate down-sampled luma neighboring samples. The method further includes predicting chroma samples for the block using cross-component linear model (CCLM) prediction, where using CCLM prediction comprises predicting the chroma samples for the block based on the down-sampled luma neighboring samples generated from the MIP. The method further includes determining a prediction block for the block based on the luma samples and the chroma samples, decoding a residual block for the block, and combining the prediction block and the residual block to decode the block.

Abstract: A video decoder, encoder, and corresponding methods for processing video data for an image block and a particular reference picture index to predict the image block are disclosed that utilize adaptive weighting of reference pictures to enhance video compression, where a decoder includes a reference picture weighting factor unit for determining a weighting factor corresponding to the particular reference picture index; an encoder includes a reference picture weighting factor assignor for assigning a weighting factor corresponding to the particular reference picture index; and a method for decoding includes receiving a reference picture index with the data that corresponds to the image block, determining a weighting factor for each received reference picture index, retrieving a reference picture for each index, motion compensating the retrieved reference picture, and multiplying the motion compensated reference picture by the corresponding weighting factor to form a weighted motion compensated reference picture

Abstract: An image encoding device and a method, and an image decoding device and a method that are designed to improve encoding efficiency in IntraBC. A screen is divided into four slices (Slices #0 through #3). In a case where reference to a different slice is prohibited, the range that can be referred to from the current CTU in Slice #2 is only the decoded portion in Slice #2, and therefore, any block in Slice #1 cannot be referred to, for example. In the case of the present technology, on the other hand, decoded different slices (Slice #0 and Slice #1) are included in the referable range, and accordingly, a block in Slice #1 can be referred to from the current CTU in Slice #2.

Abstract: A method and apparatus adapts motion vector prediction for suitability to omnidirectional video. One embodiment improves handling of temporal motion vector predictors or rescaled motion vector predictors. Another embodiment is suited to spatial motion vector predictors, and another to a combination of either temporal or spatial motion vector predictors. The method analyzes a scale factor derived from, at least one of, the time index of the predictor, the time index of the reference image's predictor, the time index of a reference image's current block, and the time index of the current block. If, for example, the scale factor is greater than one, motion vector transformation is performed before motion vector rescaling. If, however, the scale factor is less than or equal to one, the motion vector rescaling is performed before motion vector transformation.

Abstract: In a method for determining a coding cost of a coding unit (CU) that is performed by a server, a CU subject to predictive coding in an intra-frame prediction mode is determined. Pixel gradient information corresponding to the CU is performed. A division predecision on the CU is made according to the pixel gradient information. A first coding cost is determined as a second coding cost of the CU in a case that a result of the division predecision on the CU is negative. The second CU is used for determining a division policy of a coding tree unit (CTU) corresponding to the CU, and the first coding cost is a coding cost obtained by performing predictive coding on the CU by using a current size of the CU as a size of a prediction unit (PU).

Abstract: Real-time, hardware-implementable Structured Similarity (SSIM)-based rate distortion optimization (RDO) techniques for video transmission are described. The disclosed techniques provide efficient application of SSIM as a distortion metric in selecting prediction modes for encoding video for transmission. A prediction mode, at a high level, specifies which previously encoded group of pixels can be utilized to encode a subsequent block of pixels in a video frame. A less compute intensive distortion metric is first used to select a subset of candidate prediction modes. Then a more compute intensive SSIM-based selection is made on the subset. By utilizing the disclosed techniques during video encoding, tradeoffs between distortion and transmission rate can be made that are more relevant to human perception.

Abstract: In one embodiment, a method includes accessing a number of pictures of an object, constructing a modeling volume for three-dimensional modeling of the object by processing the number of pictures using a machine-learning framework, where the modeling volume is associated with a number of color and opacity information that are associated with a number of regions in the modeling volume, and rendering an image of the object from a view-point using the modeling volume, where each pixel of the image is rendered by projecting a virtual ray from the view-point and through the modeling volume, determining one or more of the number of regions in the modeling volume intersected by the virtual ray, and determining a color and an opacity of the pixel based on an accumulation of the color and opacity information associated with the one or more of the number of regions intersected by the virtual ray.

Abstract: Provided is a video decoding method including obtaining a residue of a first bit-depth with respect to a current block by decoding a bitstream; when intra predicting the current block, generating a prediction block of the current block by using a block that is previously decoded at the first bit-depth and then stored in a buffer; and generating a reconstruction block of the first bit-depth by using the prediction block and the residue of the first bit-depth. When the current block is inter predicted, the video decoding method may further include generating a prediction block of a second bit-depth by using an image previously decoded at the second bit-depth, and generating the prediction block of the current block by changing the generated prediction block of the second bit-depth to the first bit-depth. The first bit-depth is higher than the second bit-depth.

Abstract: Provided is a video decoding method including determining a displacement vector per unit time of pixels of a current block in a horizontal direction or a vertical direction, the pixels including a pixel adjacent to an inside of a boundary of the current block, by using values about reference pixels included in a first reference block and a second reference block, without using a stored value about a pixel located outside boundaries of the first reference block and the second reference block; and obtaining a prediction block of the current block by performing block-unit motion compensation and pixel group unit motion compensation on the current block by using a gradient value in the horizontal direction or the vertical direction of a first corresponding reference pixel in the first reference block which corresponds to a current pixel included in a current pixel group in the current block, a gradient value in the horizontal direction or the vertical direction of a second corresponding reference pixel in the sec