Patents by Inventor Stephane Mallat
Stephane Mallat 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).
-
Patent number: 9270995Abstract: For compressing a video signal, a local multiscale transform is applied to a frame of the video signal to obtain coefficient blocks. The coefficients of each block are distributed into a plurality of coefficient groups, and for at least one of the groups, a common exponent is determined for encoding the coefficients of the group, and respective mantissas are determined for quantizing the coefficients of the group in combination with the common exponent. Coding data including each exponent determined for a coefficient group and the mantissas quantizing the coefficients of the group in combination with this exponent are stored in an external frame buffer.Type: GrantFiled: September 27, 2013Date of Patent: February 23, 2016Assignee: ZORAN (FRANCE) S.A.Inventors: Stephane Mallat, Christophe Bernard
-
Patent number: 8953875Abstract: A digital filter bank having a number J?1 of stages is disclosed. For each integer j such that 1?j?J, the j-th stage includes a plurality of filtering units (20, 21) each receiving an input signal of the j-th stage. These filtering units include a low-pass filtering unit (20) using real filtering coefficients and at least one band-pass filtering unit (21) using complex filtering coefficients. Following each band-pass filtering unit of the j-th stage, a respective modulus processing unit (25) generates a processed real signal as a function of squared moduli of complex output values of the band-pass filtering unit. The input signal of the first stage is a digital signal supplied to the digital filter bank, while for 1<j?J, the input signal of the j-th stage includes the processed real signal generated by at least one modulus processing unit of the (j?1)-th stage.Type: GrantFiled: May 26, 2011Date of Patent: February 10, 2015Assignee: Ecole PolytechniqueInventor: Stephane Mallat
-
Patent number: 8792553Abstract: A wavelet transform is applied to successive images of a video sequence to obtain wavelet coefficients for each image, and directions of regularity are estimated in association with the wavelet coefficients. Recursive weights are also determined to be associated with the wavelet coefficients. An average multiscale bandlet coefficient associated with a wavelet coefficient for a current image is computed from this wavelet coefficient, the recursive weight associated therewith and a corresponding average multiscale bandlet coefficient computed for a previous image and associated with a wavelet coefficient offset according to the direction of regularity associated with the wavelet coefficient. The average multiscale bandlet coefficients can then be processed to produce an enhanced bandlet image to which an inverse wavelet transform is applied.Type: GrantFiled: February 6, 2008Date of Patent: July 29, 2014Assignee: Zoran (France) S.A.Inventors: Stephane Mallat, Guoshen Yu
-
Publication number: 20140064380Abstract: For compressing a video signal, a local multiscale transform is applied to a frame of the video signal to obtain coefficient blocks. The coefficients of each block are distributed into a plurality of coefficient groups, and for at least one of the groups, a common exponent is determined for encoding the coefficients of the group, and respective mantissas are determined for quantizing the coefficients of the group in combination with the common exponent. Coding data including each exponent determined for a coefficient group and the mantissas quantizing the coefficients of the group in combination with this exponent are stored in an external frame buffer.Type: ApplicationFiled: September 27, 2013Publication date: March 6, 2014Applicant: ZORAN (FRANCE) S.A.Inventors: Stephane Mallat, Christophe Bernard
-
Patent number: 8620979Abstract: For subband decomposition of a d-dimensional input signal (S) into a number K of subband components (F1-F4), a filter bank has a filtering module (801) transforming the input signal (S) into 2d components including a low-frequency component (L) and 2d?1 higher-frequency components (F1), The 2d?1 higher-frequency components are oversampled, typically by a factor 2, compared to the low-frequency component. The low-frequency component can be further decomposed by means of another filtering module having a similar structure, and the process can be iterated over any number of scales. The reconstruction filter bank has a symmetric structure, with filtering modules adapted to the oversampling of the higher-frequency components. Such filter banks are well suited to various enhancement processing applied to the subband components such as thresholding, reduction of compression distortion, reduction of measurement noise, sharpness enhancement.Type: GrantFiled: December 26, 2007Date of Patent: December 31, 2013Assignee: Zoran (France) S.A.Inventor: Stéphane Mallat
-
Patent number: 8559499Abstract: For compressing a video signal, a local multiscale transform is applied to a frame of the video signal to obtain coefficient blocks. The coefficients of each block are distributed into a plurality of coefficient groups, and for at least one of the groups, a common exponent is determined for encoding the coefficients of the group, and respective mantissas are determined for quantizing the coefficients of the group in combination with the common exponent. Coding data including each exponent determined for a coefficient group and the mantissas quantizing the coefficients of the group in combination with this exponent are stored in an external frame buffer.Type: GrantFiled: October 26, 2007Date of Patent: October 15, 2013Assignee: Zoran (France) S.A.Inventors: Stéphane Mallat, Christophe Bernard
-
Publication number: 20130058567Abstract: A digital filter bank having a number J?1 of stages is disclosed. For each integer j such that 1?j?J, the j-th stage includes a plurality of filtering units (20, 21) each receiving an input signal of the j-th stage. These filtering units include a low-pass filtering unit (20) using real filtering coefficients and at least one band-pass filtering unit (21) using complex filtering coefficients. Following each band-pass filtering unit of the j-th stage, a respective modulus processing unit (25) generates a processed real signal as a function of squared moduli of complex output values of the band-pass filtering unit. The input signal of the first stage is a digital signal supplied to the digital filter bank, while for 1<j?J, the input signal of the j-th stage includes the processed real signal generated by at least one modulus processing unit of the (j?1)-th stage.Type: ApplicationFiled: May 26, 2011Publication date: March 7, 2013Applicant: ECOLE POLYTECHNIQUEInventor: Stephane Mallat
-
Patent number: 8265161Abstract: A method and apparatus for processing or compressing an n-dimensional digital signal by constructing a sparse representation which takes advantage of the signal geometrical regularity. The invention comprises a warped wavelet packet transform which performs a cascade of warped subband filtering along warping grids of sampling points adapted to the signal geometry. It also comprises a bandeletisation which decorrelates the warped wavelet packet coefficients to produce a sparse representation. An inverse warped wavelet packet transform and an inverse bandeletisation reconstruct a signal from its bandelet representation. The invention comprises a compression system which quantizes and codes the bandelet representation, a decompression system, a restoration system which enhances a signal by filtering its bandelet representation, and a feature vector extraction system for pattern recognition applications of a bandelet representation.Type: GrantFiled: January 14, 2010Date of Patent: September 11, 2012Assignee: Zoran (France)Inventors: Christophe Bernard, Jerome Kalifa, Erwan Le Pennec, Stephane Mallat
-
Patent number: 8229245Abstract: Video processing method and means for enhancing a video stream, by computing transform coefficients using a spatio-temporal transform comprising a spatial subband transform and a causal time wavelet transform performing filterings with multiscale causal wavelets, modifying the transform coefficients with a nonlinear processing, and computing a processed video stream from the modified transform coefficients using a spatio-temporal reconstruction transform comprising an inverse subband transform and a short delay inverse time wavelet transform, where the short delay inverse time wavelet transform is implemented with wavelet filters modified with window functions to control the processing delay of the entire video processing method.Type: GrantFiled: October 10, 2006Date of Patent: July 24, 2012Assignee: Zoran (France)Inventors: Stephane Mallat, Christophe Bernard, Joan Bruna, Nicolas Laveau
-
Patent number: 8189939Abstract: A method and apparatus for enhancing a d-dimensional digital signal by taking advantage of regular geometrical structures. The invention comprises a multiscale redundant grouping lifting (102), which iteratively computes grouping arrays and bandelet signals with a redundant grouping lifting. A second step makes an enhancement (103) of bandelet signals. A third step is an inverse multiscale redundant grouping lifting (104) which iteratively updates an enhanced average signal with an inverse redundant grouping lifting that takes in input enhanced bandelet signals. It outputs an enhanced signal. The invention also comprises a subband grouping bandelet enhancement which incorporates a subband decomposition and reconstruction. It includes a super-resolution process which increases the signal resolution by computing an output enhanced signal on an output sampling grid that is larger than the input sampling grid.Type: GrantFiled: November 25, 2005Date of Patent: May 29, 2012Assignee: Zoran FranceInventor: Stephane Mallat
-
Patent number: 8054380Abstract: A method and apparatus for scaling video images, which applies more particularly to techniques for converting video sequence formats and deinterlacing. The invention comprises a parameterized interpolation (101) which computes interpolations at each new point of the output. An insensitive interval estimation (102) computes an interval where the probability distribution of the output image value is estimated to have its maximum, and an insensitive loss calculation (103) derives for each interpolation value a negative log likelihood measurement from an interval-insensitive probability distribution. After a spatial regularization (104) the invention includes a minimum loss selection (105) which selects the minimum loss interpolation at each new point and provides it as an output image value. The invention also includes a total variation increment module and a directional matching error module that can penalize the insensitive loss, for the minimum loss selection.Type: GrantFiled: April 12, 2006Date of Patent: November 8, 2011Assignee: Zoran (France)Inventors: Joan Bruna, Stephane Mallat
-
Patent number: 7944974Abstract: A method and apparatus for processing or compressing an n-dimensional digital signal by constructing a sparse representation which takes advantage of the signal geometrical regularity. The invention comprises a warped wavelet packet transform which performs a cascade of warped subband filtering along warping grids of sampling points adapted to the signal geometry. It also comprises a bandeletisation which decorrelates the warped wavelet packet coefficients to produce a sparse representation. An inverse warped wavelet packet transform and an inverse bandeletisation reconstruct a signal from its bandelet representation. The invention comprises a compression system which quantizes and codes the bandelet representation, a decompression system, a restoration system which enhances a signal by filtering its bandelet representation, and a feature vector extraction system for pattern recognition applications of a bandelet representation.Type: GrantFiled: December 17, 2002Date of Patent: May 17, 2011Assignee: Zoran (France)Inventors: Christophe Bernard, Jérôme Kalifa, Erwan Le Pennec, Stéphane Mallat
-
Publication number: 20110002384Abstract: A wavelet transform is applied to successive images of a video sequence to obtain wavelet coefficients for each image, and directions of regularity are estimated in association with the wavelet coefficients. Recursive weights are also determined to be associated with the wavelet coefficients. An average multiscale bandlet coefficient associated with a wavelet coefficient for a current image is computed from this wavelet coefficient, the recursive weight associated therewith and a corresponding average multiscale bandlet coefficient computed for a previous image and associated with a wavelet coefficient offset according to the direction of regularity associated with the wavelet coefficient. The average multiscale bandlet coefficients can then be processed to produce an enhanced bandlet image to which an inverse wavelet transform is applied.Type: ApplicationFiled: February 6, 2008Publication date: January 6, 2011Inventors: Stephane Mallat, Guoshen Yu
-
Publication number: 20100253447Abstract: For subband decomposition of a d-dimensional input signal (S) into a number K of subband components (F1-F4), a filter bank has a filtering module (801) transforming the input signal (S) into 2d components including a low-frequency component (L) and 2d?1 higher-frequency components (F1), The 2d?1 higher-frequency components are oversampled, typically by a factor 2, compared to the low-frequency component. The low-frequency component can be further decomposed by means of another filtering module having a similar structure, and the process can be iterated over any number of scales. The reconstruction filter bank has a symmetric structure, with filtering modules adapted to the oversampling of the higher-frequency components. Such filter banks are well suited to various enhancement processing applied to the subband components such as thresholding, reduction of compression distortion, reduction of measurement noise, sharpness enhancement.Type: ApplicationFiled: December 26, 2007Publication date: October 7, 2010Inventor: Stéphane Mallat
-
Publication number: 20100220783Abstract: For compressing a video signal, a local multiscale transform is applied to a frame of the video signal to obtain coefficient blocks. The coefficients of each block are distributed into a plurality of coefficient groups, and for at least one of the groups, a common exponent is determined for encoding the coefficients of the group, and respective mantissas are determined for quantizing the coefficients of the group in combination with the common exponent. Coding data including each exponent determined for a coefficient group and the mantissas quantizing the coefficients of the group in combination with this exponent are stored in an external frame buffer.Type: ApplicationFiled: October 26, 2007Publication date: September 2, 2010Inventors: Stéphane Mallat, Christophe Bernard
-
Publication number: 20100135400Abstract: A method and apparatus for processing or compressing an n-dimensional digital signal by constructing a sparse representation which takes advantage of the signal geometrical regularity. The invention comprises a warped wavelet packet transform which performs a cascade of warped subband filtering along warping grids of sampling points adapted to the signal geometry. It also comprises a bandeletisation which decorrelates the warped wavelet packet coefficients to produce a sparse representation. An inverse warped wavelet packet transform and an inverse bandeletisation reconstruct a signal from its bandelet representation. The invention comprises a compression system which quantizes and codes the bandelet representation, a decompression system, a restoration system which enhances a signal by filtering its bandelet representation, and a feature vector extraction system for pattern recognition applications of a bandelet representation.Type: ApplicationFiled: January 14, 2010Publication date: June 3, 2010Applicant: ZORAN (France)Inventors: Christophe Bernard, Jérôme Kalifa, Erwan Le Pennec, Stéphane Mallat
-
Publication number: 20100104210Abstract: Video processing method and means for enhancing a video stream, by computing transform coefficients using a spatio-temporal transform comprising a spatial subband transform and a causal time wavelet transform performing filterings with multiscale causal wavelets, modifying the transform coefficients with a nonlinear processing, and computing a processed video stream from the modified transform coefficients using a spatio-temporal reconstruction transform comprising an inverse subband transform and a short delay inverse time wavelet transform, where the short delay inverse time wavelet transform is implemented with wavelet filters modified with window functions to control the processing delay of the entire video processing method.Type: ApplicationFiled: October 10, 2006Publication date: April 29, 2010Inventors: Stephane Mallat, Christophe Bernard, Joan Bruna, Nicolas Laveau
-
Publication number: 20090278981Abstract: A method and apparatus for scaling video images, which applies more particularly to techniques for converting video sequence formats and deinterlacing. The invention comprises a parameterized interpolation (101) which computes interpolations at each new point of the output. An insensitive interval estimation (102) computes an interval where the probability distribution of the output image value is estimated to have its maximum, and an insensitive loss calculation (103) derives for each interpolation value a negative log likelihood measurement from an interval-insensitive probability distribution. After a spatial regularization (104) the invention includes a minimum loss selection (105) which selects the minimum loss interpolation at each new point and provides it as an output image value. The invention also includes a total variation increment module and a directional matching error module that can penalize the insensitive loss, for the minimum loss selection.Type: ApplicationFiled: April 12, 2006Publication date: November 12, 2009Applicant: ZORAN (FRANCE)Inventors: Joan Bruna, Stephane Mallat
-
Publication number: 20090175552Abstract: A method and apparatus for enhancing a d-dimensional digital signal by taking advantage of regular geometrical structures. The invention comprises a multiscale redundant grouping lifting (102), which iteratively computes grouping arrays and bandelet signals with a redundant grouping lifting. A second step makes an enhancement (103) of bandelet signals. A third step is an inverse multiscale redundant grouping lifting (104) which iteratively updates an enhanced average signal with an inverse redundant grouping lifting that takes in input enhanced bandelet signals. It outputs an enhanced signal. The invention also comprises a subband grouping bandelet enhancement which incorporates a subband decomposition and reconstruction. It includes a super-resolution process which increases the signal resolution by computing an output enhanced signal on an output sampling grid that is larger than the input sampling grid.Type: ApplicationFiled: November 25, 2005Publication date: July 9, 2009Inventor: Stephane Mallat
-
Patent number: 7519477Abstract: A method for determining impedance coefficients of a seismic trace comprises determining reflection coefficients of the seismic trace, for example using a sparse spike inversion, integrating the reflection coefficients with respect to time to obtain impedance coefficients, and filtering the impedance coefficients by applying a low-cut window filter. The window size and/or shape may be defined by a variable parameter which may be either specified by a user or optimized on the basis of a lateral variability parameter calculated for different values of the window parameter.Type: GrantFiled: October 20, 2006Date of Patent: April 14, 2009Assignee: BHP Billiton Innovation Pty Ltd.Inventors: Michael Edwin Glinsky, Jerome Kalifa, Stephane Mallat