Google Inc.

Pascal Massimino

Vincent Rabaud

Shumeet Baluja

Michele Mae Covell

Sergey Ioffe

Paul Gordon Wilkins

Jyrki Antero Alakuijala

Andrew Ian Russell

David Marwood

Richard Rabbat

Feng Yang

Nicholas M Johnston

Jingyu Cui

Ronald Sebastiaan Bultje

Lode Vandevenne

Jeremy Doig

Maryla Ustarroz-Calonge

Abraham Stephens

Venkatesan Esakki

Haoran Man

Yannis Guyon

Cecilia Rabess

Decoding a bitset, each bit of the bitset corresponding to a respective value in a range of a minimum value to a maximum value, includes decoding, from a compressed bitstream, indexes of bits of the bitset, each bit of the bits having a first value. Decoding the bitset also includes setting all other bits of the bitset not decoded from the compressed bitstream to a second value. Decoding the indexes of bits of the bitset includes decoding a number of the indexes of the bits of the bitset, decoding a first index of the indexes in a first range having a first lower bound and a first upper bound, and decoding a last index of the indexes in a second range having a second lower bound and a second upper bound.

Dynamic bitset coding

A sparse matrix representation of image or video data for encoding or decoding uses a boundary of non-zero coefficients within the image or video data. A bounding box encloses each non-zero coefficient within an image or video block. The coefficients enclosed within the bounding box are encoded to a bitstream along with dimensional information usable to identify the bounding box within the image or video block during decoding. Coefficients not enclosed within the bounding box are not specifically encoded within the bitstream. The dimensional information represents one or more of a shape, size, or position within the image or video block of the bounding box. The bounding box may be identified according to a scan order used to process the coefficients within the image or video block. The bounding box may be rectangular or non-rectangular.

Sparse matrix representation using a boundary of non-zero coefficients

Alpha channel post processing in image coding can include decoding, from multiple color channels of a bitstream, color channel values for an encoded image, decoding, from an alpha channel of the bitstream, alpha channel values for the encoded image, determining a bilateral filter based on a level of compression for encoding the alpha channel, post processing the alpha channel values by filtering the alpha channel values using the bilateral filter to obtain filtered alpha channel values, and generating at least a portion of a reconstructed image corresponding to the encoded image using the filtered alpha channel values and the color channel values.

Alpha channel post processing in image coding

Updating cumulative distribution functions (CDFs) during arithmetic encoding can be a challenge because the final element of the CDF should remain fixed during the update calculations. If the probabilities were floating-point numbers, this would not be too much of a challenge; nevertheless, the probabilities and hence the CDFs are represented as integers to take advantage of infinite-precision arithmetic. Some of these difficulties may be alleviated by introducing a 'mixing' CDF along with the active CDF being updated; the mixing CDF provides nonlocal context for updating the CDF due to the introduction of a particular symbol in the encoding. Improved techniques of performing arithmetic encoding include updating the CDF using two, one-dimensional mixing CDF arrays: a symbol-dependent array and a symbol-dependent array. The symbol-dependent array is a sub array of a larger, fixed array such that the sub array selected depends on the symbol being used.

Efficient update of cumulative distribution functions for image compression

A data stream accessor seeks to simulate an uninterrupted bitstream for a decoder during network interruptions. The accessor performs a set of operations including receiving an available amount of image data and storing a current execution context of the call site, passing execution flow to the decoder, and performing a subset of operations. The subset includes receiving an indication of a desired amount of image data from the decoder. If a total amount of image data at the accessor is greater than or equal to the desired amount, the desired amount is transmitted to the decoder for decoding. Otherwise, and unless an error issues, a current execution context of the decoder is stored, execution flow is passed from the decoder to the call site, and the restorable execution context is loaded as the current execution context of the call site. The subset is repeated unless the image is completely decoded.

Image decoding during bitstream interruptions

A method includes compressing an image using a quality setting, determining a quality of the compressed image based on a quality metric, and determining if the quality of the compressed image is within a quality range. In response to determining the quality of the compressed image is within the quality range, store the compressed image; and in response to determining the quality of the compressed image is not within the quality range, modify the quality setting, and repeat the compressing step with the modified quality setting, both determining steps, and the applicable in response to step.

Adaptive image compression using a quality metric

An encoder system can include a pixel grid generator to receive an image having a first dimension, generate a grid having a second dimension, add a plurality of points to positions on the grid, and map a plurality of pixels of the image to the plurality of points. The encoder system can include a color module to assign a color to each of the plurality of points using a color table, a triangulation module to generate a plurality of vertices based on the plurality of points and triangulate the grid using the vertices, and a compression module to compress the vertices as a set of compressed vertex positions and a set of vertex colors.

Image compression and decompression using triangulation

Decoding image data using lexicographic coding order with floating block-partitioning includes obtaining, from an encoded bitstream, encoded data for a defined portion of a frame, generating a reconstructed frame by decoding the encoded data, and outputting the reconstructed frame for presentation to a user. Decoding the encoded data using lexicographic coding order with floating block-partitioning includes decoding, from the encoded data, block dimension data for respective blocks from the plurality of blocks in lexicographic coding order, determining block location data for the respective blocks from the plurality of blocks in lexicographic coding order, generating reconstructed block data for the respective blocks from the plurality of blocks using the block dimension data and the block location data by decoding, from the encoded data, image content data for the respective blocks from the plurality of blocks, and including the reconstructed block data in the reconstructed frame.

Image coding using lexicographic coding order with floating block-partitioning

A method includes obtaining, at one or more computing devices, an input image, applying an image sharpening process to at least one of one or more chroma components of the input image, subsampling the one or more chroma components of the input image to reduce a spatial resolution of the one or more chroma components of the input image, encoding the input image subsequent to applying the image sharpening process and subsampling, and performing at least one of storing or transmitting the input image subsequent to encoding.

Chroma blurring reduction in video and images

A method and apparatus for performing sparse coding dictionary priming are disclosed. Sparse coding dictionary priming may include iteratively training a coding dictionary, which may include a plurality of codewords or bases. Iteratively training the coding dictionary may include identifying a sampling index cardinality, identifying a portion of a video stream, decomposing the portion of the video stream, and updating the codeword based on the portion of the video stream. Decomposing the portion of the video stream may include randomly identifying a set of codewords from the plurality of codewords wherein a cardinality of the set of codewords is the sampling index cardinality and wherein the sampling index cardinality is less a cardinality of the plurality of codewords, and determining a codeword having a maximum correlation with the portion of the video stream from the set of codewords.

Sparse coding dictionary priming

A method includes obtaining, at one or more computing devices, an input image; determining a first value of a quality metric for the input image; generating a first chroma subsampled representation of the input image; and generating a reconstructed image based on the chroma subsampled representation. The method also includes determining a second value of the quality metric for the reconstructed image; determining an error value based on the first value of the quality metric and the second value of the quality metric; and generating a second chroma subsampled representation of the input image based in part on the error value.

System and method for optimized chroma subsampling

Systems, methods and computer program products for displaying image sprites, sub-regions and three dimensional images includes (but is not limited to) receiving a composite image file containing multiple individual images and an index of the multiple individual images, parsing a Uniform Resource Locator (URL) of the composite image file as a function of the index of the multiple individual images, extracting a desired individual image from the composite image file as a function of the index and displaying the extracted individual image.

Anchors for displaying image sprites, sub-regions and 3D images

A plurality of signature bits to be used to generate a watermark in a JPEG image file is determined, wherein the JPEG image file comprises at least one quantization table. A plurality of locations in the at least one quantization table is selected. A respective value associated with each of the selected plurality of locations in the at least one quantization table is changed, based on the plurality of signature bits.

Growing community of inventors

Orsay, France

Pascal Massimino