Processing Input Signals
Patent
WO2017129546A1
Owner
TU Berlin
Filing Date
January 24, 2017
Priority Date
January 26, 2016
Portfolio
Video - TU Berlin
Intro
The method described is an almost universal bit-efficient video compressing approach that can be used in various coding applications. The main goal is to reconstruct images in high quality by using a universal image coder with easy bit-level access. The idea is to search for unsteady regions in the image and to approximate the stationary regions separately and combining both of them.
Claims
1.
Method for processing an input signal (S) and generating an output signal (Sout) based on the input signal (S) , said method comprising the steps o...
- Method for processing an input signal (S) and generating an output signal (Sout) based on the input signal (S) , said method comprising the steps of:
- in an analysis stage, extracting a plurality of kernels from the input signal (S) , wherein each kernel is described by a parameter vector that is defined by a given number of extracted kernel parameters, and
- forming the output signal (Sout) based on the extracted kernel parameters. - Method of claim 1, wherein
- the extracted kernel parameters are quantized,
- the quantized kernel parameters are coded, and
- the output signal (Sout) is formed based on the coded kernel parameters. - Method of any of the preceding claims, wherein the kernels are extracted using a kernel density mixture model algorithm.
- Method of any of the preceding claims, wherein the kernel parameters are derived using an Expectation- Maximization, EM, algorithm.
- Method of any of the preceding claims, wherein the kernels are Gaussian kernels that are defined by a one- dimensional or multi-dimensional Gaussian function.
- Method of any of the preceding claims, wherein the extracted kernel parameters are transformed to another domain prior to quantizing and coding, preferably based on FFT, KLT, DCT or predictive coding operators.
- Method of any of the preceding claims, wherein said step of quantizing is an adaptive quantizing step wherein the kernel parameters are quantized differently.
- Method of any of the preceding claims, wherein the step of coding is a predictive coding
- Method of any of the preceding claims, wherein, before quantizing the kernel parameters, a sparsification approach is used to reduce the number of kernels.
- Method of any of the preceding claims, wherein one or more iteration steps are carried out before coding the quantized kernel parameters wherein each iteration step comprises:
- generating a reconstructed signal based on the quantized kernel parameters;
- comparing the input signal (S) and the reconstructed signal,
- determining the deviation between the reconstructed signal and the input signal (S),
- if the deviation exceeds a given threshold, modifying the analysis stage, preferably with respect to the number of kernels, and/or the quantizing, and repeating the iteration until the deviation reaches an accepted range, and
- continuing with the formation of the coded signal based on the current analysis stage and/or the current quantizing if the deviation has reached the accepted range. - Method of any of the preceding claims, wherein said coded signal is decoded.
- Method of claim 11, wherein a reconstructed signal is generated based on the decoded signal.
- Method of any of the preceding claims 11-12, wherein signal features are derived from the decoded signal.
- Method of any of the preceding claims 11-13, wherein an enhanced signal with increased or decreased sample resolution is derived from the decoded continuous kernel signal equation.
- Method of any of the preceding claims, wherein the input signal (S) describes a two-dimensional picture.
- Method of claim 15, wherein each parameter vector comprises:
- a kernel weighting factor
- at least two kernel parameters that define a position of an ellipse in the two-dimensional picture,
- at least one kernel parameter which defines the
orientation of the ellipse,
- at least one kernel parameter which defines the size of the ellipse, and
- at least one kernel parameter which defines the color and/or the average intensity of the ellipse. - Method of any of the preceding claims 1-14, wherein the input signal (S) describes a video sequence.
- Method of any of the preceding claims 1-14, wherein the input signal (S) is a one- or multi-dimensional audio or speech signal.
- Method of any of the preceding claims 1-14, wherein the input signal (S) is a multi-dimensional lightfield image or lightfield video signal including multiple angular dimensions.
- Method of any of the preceding claims 11-12, wherein signal features are derived from the decoded signal.
- Encoder (100) configured to process an input signal (S) and generate a coded signal based on the input signal (S) , the encoder (100) carrying out the steps of:
- in an analysis stage, extracting a plurality of kernels from the input signal (S) , each kernel being defined by a given number of extracted kernel parameters,
- quantizing the extracted kernel parameters,
- coding the kernel parameters, and
- forming said coded signal based on the coded kernel parameters. - Decoder (200) configured to decode a coded signal that has been encoded according to any of the preceding claims, said decoder (200) being configured to carry out a signal reconstruction method for reconstruction of the encoder's (100) input signal (S), the reconstruction method comprising the steps of:
- deriving from each decoded kernel parameter set an tridimensional interpolation function, - deriving from each decoded kernel parameter set an tridimensional window function, and
- reconstructing each sample of the encoder (100) input tridimensional signal based on a weighted combination of all interpolation and window functions, or a subset thereof. - Datastream comprising coded kernel parameters that have been extracted from an input signal (S).
Abstract
An embodiment of the invention relates to a method for processing an input signal (S) and generating an output signal (Sout) based on the input signal...
An embodiment of the invention relates to a method for processing an input signal (S) and generating an output signal (Sout) based on the input signal (S), said method comprising the steps of: In an analysis stage, extracting a plurality of kernels from the input signal (S), wherein each kernel is described by a parameter vector that is defined by a given number of extracted kernel parameters, and forming the output signal (Sout) based on the extracted kernel parameters.
read more
Write Review
