1.K-SVD: An Algorithm for Designing Overcomplete Dictionaries for Sparse Representation
M.Aharon,M.Elad and A.Bruckstein, IEEE transactions on signal processing,vol.54.no.11,nov 2006
The goal of this algorithm is to find a dictionary that can yield sparse representation for the training signals. Dictionaries can be designed either by selecting from a prespecified set of transforms or by adapting the dictionary to the training signals. Designing a dictionary includes two steps, first finding the coefficients for a given dictionary and the second step includes the dictionary update with the known and fixed coefficients. It is believed that these dictionaries can outperform the commonly used predetermined dictionaries.
The K-SVD is an iterative algorithm that alternates between the sparse coding of signals that is based on the current dictionary and update of dictionary atoms to fit the data. This algorithm is flexible with any pursuit method.
For the given signal Y and dictionary D , sparse coding computes the representation coefficients X, this is done by the pursuit or greedy algorithms like MP(Matching pursuit) and OMP(orthogonal Matching Pursuit) which selects the dictionary atoms sequentially and finds an approximate solution. This approach shows good results in applications like filling up the missing pixels in an image and in image compression .
The K-SVD is an iterative algorithm that alternates between the sparse coding of signals that is based on the current dictionary and update of dictionary atoms to fit the data. This algorithm is flexible with any pursuit method.
For the given signal Y and dictionary D , sparse coding computes the representation coefficients X, this is done by the pursuit or greedy algorithms like MP(Matching pursuit) and OMP(orthogonal Matching Pursuit) which selects the dictionary atoms sequentially and finds an approximate solution. This approach shows good results in applications like filling up the missing pixels in an image and in image compression .