Optimizing Shrinkage Curves and Application in Image Denoising

Algorithm 1

Training the coefficients .

Input: paired images,

()   Initialize the table ;

()   for  

()    Initialize a parameter ; // of size .

()    for  

()      Initialize ;

()      Initialize ; 

()      Get a paired images, ;

()      Based on Definition 1, extract all patches from image and build the patch-set

()      for each patch in   do

()       Based on Definition 2, obtain a ROSM ;

()       Obtain the matrix corresponding to ;

()       Singular value decomposition, ;

()       Map to a vector, ;

()       Map to a diagonal matrix, ;

()       Map to a diagonal block matrix , according to Eq. (12) and (13);

()       Accumulation, ;

()       Accumulation, ;

()     end for

()     Obtain a optimized parameter, ;

()     if    do

()        for each patch   in   do

()         Obtain the estimation ;

()         Plug into the image canvas of the noisy image ;

()        end for

()        Obtain a new the pixels for fixed position in the image canvas;

()     end if

()     Save the to Table ;

()    end for

()    end for

Output: Table that containing the parameters .