Convolutional Sparse Coding

Basic Usage

Greyscale Images

• Convolutional sparse coding (ADMM solver)
• Convolutional sparse coding using the parallel ADMM solver
• Convolutional sparse coding using the CUDA solver
• Convolutional sparse coding (PGM solver)
• Convolutional sparse coding (constrained data fidelity)
• Convolutional sparse coding (constrained penalty term)
• Convolutional sparse coding with gradient penalty using the CUDA solver
• Convolutional sparse coding with lateral inhibition
• Convolutional sparse coding with weighted lateral inhibition

Colour Images

• Convolutional sparse coding of a colour image with a colour dictionary
• Convolutional sparse coding of a colour image with a colour dictionary (PGM solver)
• Convolutional sparse coding of a colour image with a greyscale dictionary
• Convolutional sparse coding of a colour image with a greyscale dictionary and a joint sparsity term
• Convolutional sparse coding of a colour image with a product dictionary

Image Restoration Applications

Denoising (Gaussian White Noise)

• Remove Gaussian white noise from a greyscale image using convolutional sparse coding
• Remove Gaussian white noise from a colour image using convolutional sparse coding

Denoising (Impulse Noise)

• Remove salt & pepper noise from a colour image using convolutional sparse coding with a colour dictionary
• Remove salt & pepper noise from a colour image using convolutional sparse coding with an l1 data fidelity term and an l2 gradient term, with a colour dictionary
• Remove salt & pepper noise from a hyperspectral image using convolutional sparse coding with an l1 data fidelity term and an l2 gradient term, with a dictionary consisting of the product of a convolutional dictionary for the spatial axes and a standard dictionary for the spectral axis
• Remove salt & pepper noise from a hyperspectral image using convolutional sparse coding with an l1 data fidelity term and an l2 gradient term, with a dictionary consisting of the product of a convolutional dictionary for the spatial axes and a PCA basis for the spectral axis

Inpainting

• Inpainting of randomly distributed pixel corruption with lowpass image components handled via non-linear filtering (greyscale image)
• Inpainting of randomly distributed pixel corruption with lowpass image components handled via gradient regularisation of an impulse dictionary filter (greyscale image)
• Inpainting of randomly distributed pixel corruption (greyscale image)
• Inpainting of randomly distributed pixel corruption (greyscale image) using the parallel ADMM solver
• Inpainting of randomly distributed pixel corruption (colour image)