Video super-resolution via sparse combinations of key-frame patches in a compression context
read more
Citations
Training end-to-end video processes
Offline training of hierarchical algorithms
Enhancing visual data using and augmenting model libraries
Generative methods of super resolution
Enhancement of Visual Data
References
Overview of the High Efficiency Video Coding (HEVC) Standard
Image Super-Resolution Via Sparse Representation
Example-based super-resolution
Low-complexity single-image super-resolution based on nonnegative neighbor embedding
Super-resolution through neighbor embedding
Related Papers (5)
Example-based learning for single-image super-resolution and JPEG artifact removal
Frequently Asked Questions (8)
Q2. What is the unit used in a coding-based SR procedure?
The patch is also the unit used in the SR procedure: the input image is divided into patches, preferably overlapping; the output of the algorithm is equally a set of patches, which are finally assembled to form the HR output image.
Q3. How can the authors compute neighbors with fast neighbor search algorithms?
no motion estimation is required and since the dictionary is unique for all the patches of an intermediate frame, the authors can compute the neighbors all at once with fast neighbor search algorithms.
Q4. How is the weight vector xit computed?
In NE, a sparse coding is performed in two steps: first, the support of the weight vector is found via nearest neighbor search; secondly, the coefficients are computed by solving a least squares approximation problem.
Q5. What is the advantage of the SR approach?
When adding also inter coding (random-access profile of HEVC), however, the advantage of the SR approach is lost, since slight improvements for low bit-rates are visible only for certain video sequences.
Q6. What is the definition of an internal dictionary?
In [9], e.g. an internal dictionary is built by sampling patches from a pyramid of recursively scaled images, where the LR input image itself is the “top” of the pyramid.
Q7. What is the definition of a dictionary?
The authors speak about D as an “external dictionary” when the patches that compose it are conveniently derived from a set of coupled training images; the authors call instead D an “internal dictionary” when the patches that compose it are strictly related to the content of the LR input image the authors want to super-resolve.
Q8. What is the definition of a coding-based SR procedure?
Once the dictionary is created, and the two matrices of LR and HR patch vectors, respectively Xd and Yd, arestored, the coding-based SR procedure can be then summarized as follows.