Patents by Inventor Marta Mrak
Marta Mrak has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).
-
Publication number: 20230062509Abstract: An intra-prediction mode is provided for chroma data, in which a neural network implemented attention module directs the encoding of a block of chroma data with respect to luma data for the collocated luma block.Type: ApplicationFiled: November 19, 2020Publication date: March 2, 2023Inventors: MARC GORRIZ BLANCH, MARTA MRAK, SAVERIO BLASI
-
Publication number: 20220377342Abstract: Intra-prediction modes are defined in that, predictions obtained in accordance with those modes are based on a sum of a first component dependent on a set of reference samples and a second component not dependent on the reference samples.Type: ApplicationFiled: October 2, 2020Publication date: November 24, 2022Inventors: MARIA CLAUDIA SANTAMARIA GOMEZ, SAVERIO BLASI, MARTA MRAK, EBROUL IZQUIERDO
-
Patent number: 11317102Abstract: The present invention relates to a method of decoding a video bitstream, the method comprising the steps of: receiving a bitstream representing: residual samples produced by subtracting encoder filtered motion compensated prediction samples from image samples; and motion vectors used in forming the motion compensated prediction samples; the encoder filtering process conducted on the motion compensated prediction samples at an encoder having at least one parameter; using said motion vectors to provide motion compensated prediction samples from a previously reconstructed image; decoder filtering said motion compensated prediction samples in accordance with said at least one parameter; and adding said filtered motion compensated prediction samples to said residual samples to reconstruct images. A system and apparatus corresponding to this method are also disclosed.Type: GrantFiled: March 20, 2019Date of Patent: April 26, 2022Assignee: British Broadcasting CorporationInventors: Matteo Naccari, Marta Mrak, Saverio Blasi, Andre Seixas Dias
-
Patent number: 11134242Abstract: An efficient high quality video coding and decoding technique is described. A prediction block is generated that contains modified coefficients. The modified coefficients improve the efficiency of the compression process once the prediction block is subtracted from the original block to generate a residual signal for quantisation and encoding. A similar process operates at the decoder side. The processes may operate in both the frequency and the spatial domain.Type: GrantFiled: November 4, 2014Date of Patent: September 28, 2021Assignee: BRITISH BROADCASTING CORPORATIONInventors: Marta Mrak, Ebroul Izquierdo, Saverio Giovanni Blasi
-
Publication number: 20210168378Abstract: The present invention relates to a method of decoding a video bitstream, the method comprising the steps of: receiving a bitstream representing: residual samples produced by subtracting encoder filtered motion compensated prediction samples from image samples; and motion vectors used in forming the motion compensated prediction samples; the encoder filtering process conducted on the motion compensated prediction samples at an encoder having at least one parameter; using said motion vectors to provide motion compensated prediction samples from a previously reconstructed image; decoder filtering said motion compensated prediction samples in accordance with said at least one parameter; and adding said filtered motion compensated prediction samples to said residual samples to reconstruct images. A system and apparatus corresponding to this method are also disclosed.Type: ApplicationFiled: March 20, 2019Publication date: June 3, 2021Applicant: British Broadcasting CorporationInventors: Matteo NACCARI, Marta MRAK, Saverio BLASI, Andre Seixas DIAS
-
Patent number: 10972747Abstract: A method of encoding image or video content within a set of targets such as a time, a complexity, quality, and bitrate, using an encoder having a plurality of coding configurations, comprises setting an initial coding configuration in the encoder, where coding configurations differ one from the other in the number of options tested for each of one or more tools. A first part of the content is encoded in the initial configuration and measures are made of the number of times a specified tool is used and a representative time taken to test an option for that specified tool. From those measures a prediction is made of the time difference between the time taken to encode content using the initial coding configuration and the time taken to encode content using another coding configuration in which a different number of options are tested for that specified tool.Type: GrantFiled: February 19, 2018Date of Patent: April 6, 2021Assignee: British Broadcasting CorporationInventors: Marta Mrak, Saverio Blasi, Ebroul Izquierdo, shenglan Huang
-
Patent number: 10893265Abstract: In video coding or decoding, coding blocks in a current frame are predicted from blocks in previously encoded frames with prediction samples and associated information organised in a prediction unit. A decision on which prediction unit type to consider for coding is made upon analysis of the distribution of a sum of absolute difference or other function of the residual signal over the block.Type: GrantFiled: November 27, 2015Date of Patent: January 12, 2021Assignee: British Broadcasting CorporationInventors: Matteo Naccari, Andrea Gabrieliini, Marta Mrak
-
Publication number: 20200007884Abstract: A method of encoding image or video content within a set of targets such as a time, a complexity, quality, and bitrate, using an encoder having a plurality of coding configurations, comprises setting an initial coding configuration in the encoder, where coding configurations differ one from the other in the number of options tested for each of one or more tools. A first part of the content is encoded in the initial configuration and measures are made of the number of times a specified tool is used and a representative time taken to test an option for that specified tool. From those measures a prediction is made of the time difference between the time taken to encode content using the initial coding configuration and the time taken to encode content using another coding configuration in which a different number of options are tested for that specified tool.Type: ApplicationFiled: February 19, 2018Publication date: January 2, 2020Inventors: Marta Mrak, Saverio Blasi, Ebroul Izquierdo, shenglan Huang
-
Patent number: 10298943Abstract: In video coding, where differences between input picture values and picture prediction values are transformed in a block based transform, the differences are formed in a series of parallel steps. A first step conducted in parallel upon a first subset of pixels uses prediction values based wholly on previously processed blocks. This first subset can include anchor pixels which are not contiguous with any previously processed block. A second step conducted in parallel upon a second subset includes pixels which are predicted from pixels of the first subset.Type: GrantFiled: December 23, 2011Date of Patent: May 21, 2019Assignee: BRITISH BROADCASTING CORPORATIONInventors: Marta Mrak, Thomas Davies, David John Flynn, Andrea Gabriellini
-
Patent number: 9877025Abstract: A video decoder has a transform mode and a transform skip mode and in the transform skip mode a residual prediction mode is selected. A compressed bit-stream is entropy-decoded to obtain a block of quantised values. An inverse quantiser operates in the transform mode to form transform coefficients and in the transform skip mode to form differential residual values. In the transform mode an inverse transform is performed to form residual values. In the transform skip mode residual prediction is used perform residual prediction using high accuracy locally decoded residual value. The residual value is then down-scaled to obtain the reconstructed residual. A block predictor is formed from decoded image values at the same precision at which video samples are stored in the frame memory.Type: GrantFiled: July 12, 2013Date of Patent: January 23, 2018Assignee: British Broadcasting CorporationInventors: Matteo Naccari, Marta Mrak
-
Publication number: 20170347092Abstract: In video coding or decoding, coding blocks in a current frame are predicted from blocks in previously encoded frames with prediction samples and associated information organised in a prediction unit. A decision on which prediction unit type to consider for coding is made upon analysis of the distribution of a sum of absolute difference or other function of the residual signal over the block.Type: ApplicationFiled: November 27, 2015Publication date: November 30, 2017Inventors: Matteo Naccari, Andrea Gabriellini, Marta Mrak
-
Publication number: 20160269725Abstract: An efficient high quality video coding and decoding technique is described. A prediction block is generated that contains modified coefficients. The modified coefficients improve the efficiency of the compression process once the prediction block is subtracted from the original block to generate a residual signal for quantisation and encoding. A similar process operates at the decoder side. The processes may operate in both the frequency and the spatial domain.Type: ApplicationFiled: November 4, 2014Publication date: September 15, 2016Inventors: Marta MRAK, Ebroul IZQUIERDO, Saverio Giovanni BLASI
-
Publication number: 20160073124Abstract: For a video sequence including a composite image having a foreground image and a transparency mask, a video encoder determines whether the transparency mask is the same as a transparency mask of a preceding image. Where the transparency mask is not the same, the transparency mask is encoded as an image. The encoder then transmits the encoded foreground image, any encoded transparency mask and a flag signifying whether the encoded transparency mask for a preceding image is to be used in association with the encoded foreground image of the current image.Type: ApplicationFiled: March 31, 2014Publication date: March 10, 2016Inventors: Marta Mrak, Matteo Naccari
-
Publication number: 20150245030Abstract: In video encoding, where image values are transformed to provide transform coefficients and transform coefficients are quantised to derive a compressed bitstream, an ID profile is selected at the encoder and used with a representative image intensity to scale image values in the block. The selected ID profile is signalled to the decoder with the compressed bitstream. At the decoder, the ID profile is extracted from the compressed bitstream and—after performing inverse quantisation and an inverse transform to form image values—those image values are inverse scaled.Type: ApplicationFiled: June 19, 2013Publication date: August 27, 2015Inventors: Marta Mrak, Matteo Naccari, David Flynn, Andrea Gabriellini
-
Patent number: 9094688Abstract: In the encoding of video data, a list of references pictures is constructed to be used in decoding. Prediction (direction) modes are defined by n-tuples (x0, x1, . . . , xM-1), where xm specifies prediction option of the m-th component of the motion vector field, m=0, . . . , M-1, where M components to the motion fields exist; where a coding option xm is an element from a set composed of all defined combinations of motion prediction modes and reference picture indices, with the addition of an option “/”, xm?{“/”, 0, 1, . . . , rm?1}; where option “/” specifies that the component is not used, and other options 0, 1, . . . , rm?1 specify one of rm combinations. A subset of all possible n-tuples is provided and a prediction mode used for predicting an inter-coded block is defined by reference to that subset.Type: GrantFiled: June 25, 2012Date of Patent: July 28, 2015Assignee: British Broadcasting CorporationInventors: Nikola Sprljan, Marta Mrak
-
Publication number: 20150016515Abstract: A video decoder has a transform mode and a transform skip mode and in the transform skip mode a residual prediction mode is selected. A compressed bit-stream is entropy-decoded to obtain a block of quantised values. An inverse quantiser operates in the transform mode to form transform coefficients and in the transform skip mode to form differential residual values. In the transform mode an inverse transform is performed to form residual values. In the transform skip mode residual prediction is used perform residual prediction using high accuracy locally decoded residual value. The residual value is then down-scaled to obtain the reconstructed residual. A block predictor is formed from decoded image values at the same precision at which video samples are stored in the frame memory.Type: ApplicationFiled: July 12, 2013Publication date: January 15, 2015Inventors: Matteo Naccari, Marta Mrak
-
Patent number: 8923406Abstract: Video encoding or decoding utilizing a spatial transform operating on rows and columns of a block, with a set of transform skip modes including: transform on rows and columns; transform on rows only; transform on columns only; no transform. An indication of the selected mode is provided to the decoder. Coefficients are scaled by a factor dependent upon the norm of the transform vector of the skipped transform to bring the untransformed image values to the same level as transformed coefficients.Type: GrantFiled: November 4, 2013Date of Patent: December 30, 2014Assignee: British Broadcasting CorporationInventors: Marta Mrak, Andrea Gabriellini, Nikola Sprljan, David Flynn
-
Publication number: 20140064379Abstract: In video encoding, a video input received in 4:2:2 is resampled. Residuals are formed through the use of reference samples stored in the native 4:2:2, before transforming, quantising and entropy coding to form an encoded bitstream in the resampled format. The encoded bitstream contains a message indicating chrominance resampling and a selected chrominance resampling filter for the decoder to use. An encoder may have a first mode in which the 4:2:2 is up-sampled to 4:4:4 and a second mode in which the 4:2:2 is down-sampled to 4:2:0.Type: ApplicationFiled: September 5, 2013Publication date: March 6, 2014Inventors: Marta Mrak, Andrea Gabriellini
-
Publication number: 20140056362Abstract: Video encoding or decoding utilising a spatial transform operating on rows and columns of a block, with a set of transform skip modes including: transform on rows and columns; transform on rows only; transform on columns only; no transform. An indication of the selected mode is provided to the decoder. Coefficients are scaled by a factor dependent upon the norm of the transform vector of the skipped transform to bring the untransformed image values to the same level as transformed coefficients.Type: ApplicationFiled: November 4, 2013Publication date: February 27, 2014Applicant: British Broadcasting CorporationInventors: Marta Mrak, Andrea Gabriellini, Nikola Sprljan, David Flynn
-
Publication number: 20130343457Abstract: In video coding, where differences between input picture values and picture prediction values are transformed in a block based transform, the differences are formed in a series of parallel steps. A first step conducted in parallel upon a first subset of pixels uses prediction values based wholly on previously processed blocks. This first subset can include anchor pixels which are not contiguous with any previously processed block. A second step conducted in parallel upon a second subset includes pixels which are predicted from pixels of the first subset.Type: ApplicationFiled: December 23, 2011Publication date: December 26, 2013Applicant: BRITISH BROADCASTING CORPORATIONInventors: Marta Mrak, Thomas Davies, David John Flynn, Andrea Gabriellini