Linear Transform-Based Image Processing Techniques

摘要

Signal processing techniques utilize operations performed in linear transform domains to perform operations including noise reduction, noise shaping, and gradient integration. The standard wavelet shrinkage method may be altered to constrain the modification of wavelet coefficients towards an initial estimate of the adjusted coefficients. The initial estimates can be computed by applying an edge detection filter to an input image. The wavelet shrinkage method may additionally be altered to include a noise preservation factor that enables an amount of noise to be preserved to avoid the production of artifacts. Integration of modified multi-dimensional gradients may also be performed in the wavelet (or other linear transform) domain by performing simple integration of the gradient (summing) separately in each dimension, obtaining a linear transform representation of the resulting signals, and combining the linear transforms.

主权项

1. A non-transitory program storage device, readable by a processor and comprising instructions stored thereon to cause one or more processors to:
apply an invertible linear transform to a first signal to obtain a transform representation of the first signal; and adjust one or more components of the transform representation of the first signal based, at least in part, on an initial estimate of the adjusted one or more components to generate a modified transform representation.