French Institute for Research in Computer Science and Automation

Rennes, France

French Institute for Research in Computer Science and Automation

Rennes, France
SEARCH FILTERS
Time filter
Source Type

Patent
University of British Columbia, University of Toronto, French Institute for Research in Computer Science and Automation | Date: 2017-02-22

A method is for estimating a three-dimensional (3D) representation of a set of two-dimensional (2D) curves of a concept drawing, the estimate of the 3D representation corresponding to a 3D object underlying the concept drawing. The method comprises: obtaining a representation of a set of 2D curves a concept drawing that represent a 3D object underlying the concept drawing; determining an energy function based on the set of 2D curves, the energy function comprising one or more terms, each term reflective of a preference for a 3D representation based on a characteristic of the 2D curves which reflects how concept drawings are commonly perceived to represent 3D objects; and performing an optimization which minimizes the energy function to thereby determine the 3D representation.


A method for determining a set of points to be supported for an object to be manufactured by means of an additive manufacturing method, characterised in that it comprises a step consisting of subdividing the object into successive layers, each layer corresponding to a thickness of material deposited during the manufacture of the object; and, for each layer, adding, to a set of points to be supported, points to be supported (Ps) on the surface of the object that make it possible to ensure the stability of all of the sub-objects (2n), a sub-object being defined as a solid resulting from the manufacture of the i first layers (Ci) of the object.


Patent
Thomson Licensing, French Institute for Research in Computer Science and Automation | Date: 2017-04-05

A decoding method is disclosed that comprises:- decoding (S110, S120) a plurality of picture blocks;- clustering (S130) the plurality of decoded picture blocks into clusters;- decoding (S140) at least one projection function for each cluster;- applying (S150), on at least one decoded picture block, the at least one projection function associated with the cluster to which said at least one decoded picture block belongs.


Patent
Thomson Licensing, French Institute for Research in Computer Science and Automation | Date: 2017-04-05

The disclosure relates to a method for predicting a current block of pixels in a current frame, comprising:- obtaining (11) a first prediction of the current block, delivering a first predicted block,- updating said first predicted block by implementing at least one iteration (12) of the following steps: obtaining (121) a current patch formed by a current template of the current block and an updated predicted block obtained at a previous iteration;o searching (122) for K nearest neighbor patches of said current patch in a decoded part of the current frame or in at least one reference frame;o determining (123) a linear combination of the K nearest neighbor patches which approximates the current patch, or of the K templates of the nearest neighbor patches which approximates the current template, according to a predetermined criterion, obtaining (124) an updated predicted block by applying said linear combination to the corresponding K blocks of the nearest neighbor patches.


Patent
Thomson Licensing, French Institute for Research in Computer Science and Automation | Date: 2017-01-11

A technique for efficient multimedia content indexing, search and retrieval is proposed based on Product Quantization (PQ) codes, by using the processor more efficiently to make the scanning faster. The technique indexes a multimedia content database according to a depth-first tree structure based on index groups, each group represented by at least one bit per sub-quantizer, and performs the search by storing in a fast access memory distance table segments for at least one index group at a time. The m b PQ code may be transformed from an original (2 m) (b/2) PQ code.


Patent
Thomson Licensing, French Institute for Research in Computer Science and Automation | Date: 2017-01-18

A method and a device for encoding and decoding High Dynamic Range (HDR) images and videos whose pixels are represented by high bit depth integers. In particular, a scalable compression scheme involving numerical representation and prediction of color information for the HDR enhancement layer given and LDR layer by generating color prediction information in response to a reconstructed LDR layer and a reconstructed HDR layer.


Patent
Thomson Licensing, French Institute for Research in Computer Science and Automation | Date: 2017-01-11

A technique for efficient multimedia content indexing, search and retrieval is proposed based on Product Quantization (PQ) codes, by using the processor more efficiently to make the scanning faster. The technique indexes a multimedia content database according to a tree structure based on index groups, each group represented by at least one bit per sub-quantizer, for less than all sub-quantizer. In addition, sub-quantizers centroids are grouped and each group, named minimum group, is represented by their minimum distance. The search is performed by storing in a fast access memory distance table segments for at least one index group and at least one minimum group at a time.


Patent
French Institute for Research in Computer Science and Automation | Date: 2017-02-22

Method of automatic synthesis of circuits comprising the generation of a network of regular processes reading or writing data in channels, according to which a single producer process is authorized to write in a channel and a single consumer process is authorized to read in a channel; and a synchronization unit associated with said channel authorizes or disables the implementation of a new iteration of said producer process, respectively consumer process, as a function of a comparison of a position of execution determined as a function of the value of a new iteration collected from the producer process, respectively consumer process, and of a position of execution determined as a function of a last iteration value collected from the consumer process, respectively producer process.


Patent
Thomson Licensing, French Institute for Research in Computer Science and Automation | Date: 2017-04-19

A method for clustering patches of a degraded version of an image is disclosed wherein the clusters are based on an error computed between a patch of the degraded version processed with an upgrade function and a patch of a full graded version of the image. In other words, the source image is used to determine clusters for the image to restore. A restoration framework based on a clustering according to the present principles is also disclosed wherein the optimal clusters are estimated from the degraded signal and aside indications obtained from the above described clustering method applied to a subset of the degraded signal for which the source signal is known.


Patent
Thomson Licensing, French Institute for Research in Computer Science and Automation | Date: 2017-04-12

A method and an apparatus for de-noising an image, and in particular, de-noising an image using video epitome based on a source video image. An embodiment of the present principles provides a method of processing an image in a video, comprising: accessing an encoded version of the image and generating a decoded version of the image; accessing a video epitome associated with the image, wherein the video epitome was extracted from a source version of the image prior to encoding; and generating a de-noised version of the image using the decoded version of the image and the video epitome. The use of the epitome from a source version of the image provides for a high quality epitome for improving the de-noising process.

Loading French Institute for Research in Computer Science and Automation collaborators
Loading French Institute for Research in Computer Science and Automation collaborators