# Highly Nonlinear Approximations for Sparse Signal Representation

# Matlab Files

## Image Processing:

- CalcPSNR returns the PSNR between the original image and its approximation
- DCos generates a matrix whos columns are discrete cosine vectors
- DetectLines returns the index of vertical impulsive lines in an image
- GenerateHats Generates a hat dictionary
- GenerateTrapezium generates a vector representing a trapezium
- ImageApproximation Returns an approximation of an image generated by choosing atoms from dictionary using either thresholding or a greedy algorithm
- RemoveSimilarAtoms removes any similar atoms from a given dictionary
- TranslatePrototype translates a vector to construct either a dictionary or a basis

All the presented routines are described in detail in our manual (ps,pdf).