Hand Depth Image Denoising and Superresolution via Noise-Aware Dictionaries

Algorithm 1

The K-SVD algorithm [19].

  input: training data , residual threshold , and iteration number

  output: a dictionary

() Initialize as overcomplete DCT dictionary;

() for    do

()   foreach    do /  sparse coding phase  /

()     Using the OMP (Algorithm 2 or Algorithm 3) to solve

       s.t. ;

()   end

()   for    do   / atom update phase /

()     Find the set of patches which use the th atom

       ;

()     foreach    do

()       ;  /   denotes the th column of   /

()     end

()     Set as the matrix whose columns are ;

()     Apply SVD decomposition ;

()     Update the th column of dictionary by the first column of ;

()     Update the coefficient values to be the first column of multiplied by ;

()   end

() end