Abstract: A method is disclosed that comprises training encoder and decoder neural networks to learn encoder and decoder parameters, wherein the method comprises, during training, quantizing and freezing learned decoder parameters decoding layer per decoding layer at different epochs.

Abstract: In end-to-end compression, a deep neural-network based encoder can be used to encode an image. The embeddings output from the encoder are quantized and encoded with a lossless encoder. Advantageously, at least one embodiment allows improving the latent entropy coding by further reducing the redundancies in the quantized latent. To that end, at least one embodiment discloses taking into account channels importance by coding an indication of a channel activity (or significance): performing post-conditional entropy coding by computing conditional probability based on a context afterwards: using channels reordering to improve inter channel correlation: or performing RDOQ like process by optimizing the main latent for a particular image.

Abstract: A method comprising obtaining (601) at least one characteristic of a current sample of a picture before reconstructing the current sample; determining (602) parameters of a film grain synthesis process to be applied to the current sample from the at least one characteristic; storing (603) the determined parameters with information representative of a location of the current sample; reconstructing (604) the current sample; and, applying (605) the film grain synthesis process on the current sample with the determined parameters.

Abstract: Apparatuses and methods are disclosed including techniques for encoding video data. Techniques disclosed include obtaining video data, including data representing a video data region, and obtaining correction value(s) that represent a correction to a cross-component model for predicting a chroma sample from the video data region based on a corresponding luma sample. Then, encoding the correction value(s) and the video data into coded video data. Apparatuses and methods are also disclosed that include techniques for decoding coded video data. Techniques disclosed include receiving coded video data, coding video data including data representing a video data region and correction value(s). Then, decoding the correction value(s) from the coded video data. The decoded correction value(s) are used to adjust the cross-component model, applied for intra-prediction of a chroma sample from the video data region.

Abstract: A method and an apparatus for selecting regions used for film grain modeling are provided. A reconstructed image is obtained from an encoding of an input image. At least one region of the reconstructed image is selected based on at least one coding parameter of the region when encoding the input image, and film grain parameters are determined from the at least one region selected and the input image.

Abstract: Systems, methods, and instrumentalities are disclosed for performing stepwise scaling conversion to piecewise linear for film grain synthesis. The device may obtain a first set of film grain scaling factors. Each of the first set of film grain scaling factors may be associated with an intensity interval. The first set of film grain scaling factors may be represented by a stepwise function. The device may generate a second set of film grain scaling factors from the first set of film grain scaling factors. The second set of film grain scaling factors may be represented by a piecewise linear function or a lookup table (LUT).

Abstract: A particular implementation determines parameters of a generative probabilistic model from visual descriptors extracted from at least one image. The extracted visual descriptors are quantized and encoded using the model-based arithmetic encoding to be stored or for transmission to a decoder. The model parameters are also stored to be available to a decoder, or transmitted directly to a decoder. A decoder uses the stored, or received, model parameters to reconstruct the generative probabilistic model and then to decode the visual descriptors. The visual descriptors are used for image analysis tasks, such as image retrieval or object detection. A particular implementation uses a Gaussian mixture model as a generative probabilistic model.

Abstract: The present disclosure relates to a method for selecting at least one item, from a plurality of items described by a plurality of attributes, accounting for a value and a weighting factor assigned to one of the attributes. According to at least one embodiment, the same weighting factor is assigned to the values of the same attribute for each of the items, and the method comprises: -adjusting one of the weighting factors according to an adjustment element received by a communication terminal; -selecting one of the items accounting for the adjusted weighting factor.

Abstract: A method for pairing a first device with a second device is disclosed. A salient idea is to generate a discriminant and visually differentiable visual representation from a unique identifier of a device, to make it available for example together with the unique identifier on a printed package of the device, and to regenerate it for display as part of the pairing process. A user checking the displayed visual representation matches the printed visual representation of the package can successfully terminate a pairing process.

Abstract: The application concerns a method for transmitting video data in a data stream and associated metadata, said metadata representing associated video data structure. The method comprises the steps of—determining the structure of the video data, —splitting said structure into chunks in an incremental manner, —transmitting said chunks in the data stream at time intervals before the video data it is related to. The application concerns also a method for transmitting fingerprint information associated with video data into a video data stream. The method comprises the steps of: —determining fingerprint associated with at least one frame of said video data, —inserting at least a message into the video stream comprising said fingerprint information.

Abstract: The present disclosure relates to a method, adapted to be implemented in an electronic device, for selecting a content, during a backup operation, amongst a plurality of contents coding audiovisual data and being accessible by the electronic device. According to one embodiment, the method includes obtaining, for each accessible content, one audiovisual descriptor derived from the audiovisual data, and selecting the accessible content according to one backup rule taking into account the obtained audiovisual descriptor. The present disclosure also relates to corresponding electronic device, system, computer readable program product and computer readable storage medium.

Abstract: A method for color grading a first color frame according to a first reference color frame is described. The method includes selecting a color transformation already applied to one second color frame, according to a second reference color frame, the second color frame being different from the first color frame and applying, to said first color frame, the selected color transformation. The disclosure also relates to corresponding electronic device, computer readable program product and computer readable storage medium.

Abstract: Rendering second screen information on a second screen device. A user device receives information characterizing the source of a content data stream and the content data stream; generates a watermark payload using a plurality of parameters comprising the information and a time of the user device; and inserts a watermark comprising the watermark payload into the content data stream that is rendered. The second screen device captures the watermark and extracts the watermark payload and then at least one of the plurality of parameters; generates, using the at least one extracted parameter, a request that is sent the request to a server; receives information from the server in response to the request; and displays the received information on a screen of the second screen.

Abstract: The method comprises the generation of a deteriorated image associated with a first bit rate and of a deteriorated color transform associated with a second bit rate, an evaluation of a first distortion by comparison of the image with a first test image obtained by the application of the color transform to the deteriorated image, and of a second distortion by comparison of the image with a second test image obtained by the application of the deteriorated color transform to the image, and the determination of a final conjoint bit rate associated with a best trade-off between a conjoint bit rate (R=R1+R2) and a global distortion (dE=dE1+DE2). The optimized deteriorated image and the optimized deteriorated color transform are then embedded based on this conjoint bit rate.

Abstract: A device and a method for online storage, device and method for searching for similar content, a device and a method of transmission and a device and a method. Encrypted data is saved at a provider of online services. With the encrypted data, encrypted hashing data is saved with a public key and the content to save is encrypted with the encrypted hash. This advantageously enables data duplication at the online service provider to be prevented while preserving the private life of the users of the service. In order to search for content similar to reference multimedia data, fingerprints are also saved at the service provider. In order to limit the number of false positives returned, the fingerprint can further contain a search fingerprint, an encrypted selection fingerprint.

Abstract: Isolation of an active participant in a group of participants commences by first capturing images and audio of participants. Thereafter, an active one of the participants in the group of participants (e.g., a participant that is currently speaking) is identified. After identification of the active participant, at least one of participants' images and participants' audio are rendered to isolate the active participant.

Abstract: Methods and apparatus for generating a fingerprint of an audio signal are disclosed. The method comprises: detecting peaks in a representation of a temporal spectrum of frequencies of the audio signal, a peak being defined as a point in the representation which has a higher energy than its neighboring points; and generating the fingerprint of the audio signal as a function of a distribution of positions of the detected peaks along a frequency axis and a distribution of positions of the detected peaks along a time axis. The fingerprint of the disclosure is not only robust to many types of noise, but also robust against time scale modification and frequency shifting.

Abstract: A particular implementation determines parameters of a generative probabilistic model from visual descriptors extracted from at least one image. The extracted visual descriptors are quantized and encoded using the model-based arithmetic encoding to be stored or for transmission to a decoder. The model parameters are also stored to be available to a decoder, or transmitted directly to a decoder. A decoder uses the stored, or received, model parameters to reconstruct the generative probabilistic model and then to decode the visual descriptors. The visual descriptors are used for image analysis tasks, such as image retrieval or object detection. A particular implementation uses a Gaussian mixture model as a generative probabilistic model.

Abstract: A process of hydrodesulphuration of at least one gasoline cut implementing a catalyst comprising, in its oxide form, at least one metal from group VIB and/or at least one metal from group VIII of the periodic table, present in the form of at least one polyoxometalate of the formula (HhXxMmOy)q?, wherein X is an element selected from phosphorus, silicon, boron, nickel and cobalt, M is one or more element(s) selected from molybdenum, tungsten, nickel and cobalt, h is an integer from 0-12, x is an integer from 0-4, m is an integer 5, 6, 7, 8, 9, 10, 11, 12 and/or 18, y is an integer of 17-72 and q is an integer of 1-20, the polyoxometalates being present within a mesostructured silicon oxide matrix having a pore size of 1.5-50 nm and having amorphous walls of thickness 1-30 nm, the catalyst being sulphured before use in the process.

Abstract: A method of generating a descriptor includes determining, in an image, a circular region having a center at an interest point to be described; extracting strips extending in the circular region in mutually different directions and intersecting together at the interest point; dividing each of the strips into patches; characterizing each of the patches using Haar features to acquire elements of the descriptor corresponding to the respective Haar features; and reordering the elements of the descriptor according to the strip corresponding to a maximum local.