Reduce Video Data
Patent
US8743962
Owner
TU Berlin
Filing Date
March 01, 2011
Priority Date
March 01, 2011
Portfolio
Video - TU Berlin
Intro
Video encoding methods save motion vector fields locally in a buffer to predict the following frames, thus reducing the size of the error pattern that is needed to decode the frames. If several pixel blocks have identical motion vectors, it is sufficient to save only one of these pixel blocks. The presented method reduces the amount of data, by using a model with 8 parameters of a higher order.
Claims
1.
Method for providing a compressed video bitstream related to consecutive pictures of a video sequence, wherein the pictures are defined by pixels, ...
- Method for providing a compressed video bitstream related to consecutive pictures of a video sequence, wherein the pictures are defined by pixels, said method comprising the steps of: (a) applying at least one higher-order motion model to a motion vector field and generating at least one higher-order motion parameter set for said motion vector field, wherein said motion vector field comprises a plurality of first motion vectors, each first motion vector being assigned to a pixel block and describing the local displacement of the pixel block from the preceding picture to the consecutive picture, and wherein each pixel block comprises a plurality of pixels, and wherein the higher-order motion parameter set is calculated using a Helmholtz Tradeoff estimator; (b) calculating a second motion vector for each pixel block based on the higher-order motion parameter set; (c) calculating the deviation between the first motion vector and the second motion vector for each pixel block; and (d) generating a video bitstream comprising the higher-order motion parameter set, and for each pixel block an information which indicates or is based on the deviation calculated in step (c).
- The method of claim 1 wherein said information includes either the assignment of the higher-order motion parameter set to the respective pixel block or the assignment of the first motion vector to the respective pixel block; wherein the higher-order motion parameter set is assigned to the respective pixel block if the deviation between the first motion vector and the second motion vector is below a predefined threshold; wherein the first motion vector is assigned to the respective pixel block if the deviation between the first motion vector and the second motion vector exceeds the predefined threshold; and wherein said video bitstream is generated such that it comprises the higher-order motion parameter set, and for each pixel block the corresponding first motion vector or the assignment to the higher-order motion parameter set.
- The method of claim 1 further comprising the steps of: providing a flag for each pixel block, said flag indicating the assignment of the first motion vector or the higher-order motion parameter set to the respective pixel block, and including the flags into the video bitstream and transmitting the video bitstream to a decoder.
- The method of claim 1 further comprising the steps of: applying a plurality of higher-order motion models to the motion vector field and generating a plurality of higher-order motion parameter sets for the entire motion vector field; and re-calculating a plurality of second motion vectors for each pixel block and each higher-order motion parameter set.
- The method of claim 4 further comprising the steps of: determining for each pixel block whether the deviation between the first motion vector and the corresponding second motion vector is below said predefined threshold for at least any of said higher-order motion parameter sets, and in the affirmative assigning the respective higher-order motion parameter set to the pixel block.
- The method of claim 4 wherein for each pixel block the following steps are carried out: determining for each higher-order motion parameter set the corresponding second motion vector, determining the deviation between the first motion vector and each second motion vector; out of said plurality of second motion vectors, determining the best matching second motion vector, the best matching second motion vector being the second motion vector having the smallest deviation from the first motion vector; and assigning the higher-order motion parameter set of the best matching second motion vector to the pixel block, if the deviation between the first motion vector and the best matching second motion vector is below said predefined threshold.
- The method of claim 1 wherein the pictures of said video sequence are divided into at least two sections, each section comprising a plurality of pixel blocks.
- The method of claim 7 wherein for each section an individual motion vector field is generated; wherein at least one higher-order motion model is applied to each of the motion vector fields; wherein at least one higher-order motion parameter set is generated for each motion vector field; wherein, for each section, section-related data are generated, said section-related data comprising the higher-order motion parameter set, and for each pixel block the corresponding first motion vector or the assignment to the higher-order motion parameter set; and wherein said video bitstream is generated based on the section-related data of all sections.
- The method of claim 1 wherein a warping matrix is derived as higher-order motion parameter set.
- The method of claim 9 wherein the higher-order motion parameters are compressed before including in the compressed video bitstream.
- The method of claim 1 wherein the higher-order motion parameter set is calculated based on a 4-, 6- or 8-parameter model.
- Encoder comprising a first encoder unit adapted to apply at least one higher-order motion model to a motion vector field and generating at least one higher-order motion parameter set for the said motion vector field, said motion vector field comprising a plurality of first motion vectors, each first motion vector being assigned to a pixel block and describing the local displacement of a pixel block from the preceding picture to the consecutive picture, wherein each pixel block comprises a plurality of pixels, and wherein the higher-order motion parameter set is calculated using a Helmholtz Tradeoff estimator; a second encoder unit adapted to calculate a second motion vector for each pixel block based on the higher-order motion parameter set; a third encoder unit adapted to calculate the deviation between the first motion vector and the second motion vector for each pixel block; and a fourth encoder unit adapted to generate a video bitstream comprising the higher-order motion parameter set, and for each pixel block an information which indicates or is based on the deviation calculated by the third unit.
- Encoder according to claim 12 wherein said information includes either a flag assigning the higher-order motion parameter set to the respective pixel block or a flag assigning the first motion vector to the respective pixel block; wherein said third encoder unit is adapted to assign the higher-order motion parameter set to the respective pixel block if the deviation between the first motion vector and the second motion vector is below a predefined threshold; wherein said third unit is further adapted to assign the first motion vector to the respective pixel block if the deviation between the first motion vector and the second motion vector exceeds the predefined threshold; and wherein said fourth encoder unit is adapted to include the first motion vectors into the video bitstream for all pixel blocks to which the first motion vector has been assigned by said third unit.
- Encoder according to claim 13 wherein said fourth encoder unit is adapted to re-calculate pictures based on the data contained in said video bitstream, wherein said fourth encoder unit is further adapted to calculate deviation data indicating the deviation between the re-calculated pictures and the original pictures processed by the first encoder unit, and wherein said fourth encoder unit is further adapted to include the deviation data as such or encoded into said video bitstream.
Abstract
An embodiment of the invention relates to a method for providing a compressed video bitstream related to consecutive pictures of a video sequence, whe...
An embodiment of the invention relates to a method for providing a compressed video bitstream related to consecutive pictures of a video sequence, wherein the pictures are defined by pixels, said method comprising the steps of:
(a) applying at least one higher-order motion model to a motion vector field and generating at least one higher-order motion parameter set for said motion vector field, wherein said motion vector field comprises a plurality of first motion vectors, each first motion vector being assigned to a pixel block and describing the local displacement of the pixel block from the preceding picture to the consecutive picture, and wherein each pixel block comprises a plurality of pixels;
(b) calculating a second motion vector for each pixel block based on the higher-order motion parameter set;
(c) calculating the deviation between the first motion vector and the second motion vector for each pixel block; and
(d) generating a video bitstream comprising the higher-order motion parameter set, and for each pixel block an information which indicates or is based on the deviation calculated in step (c).
read more
Write Review