Sparse Codebook Model of Local Structures for Retrieval of Focal Liver Lesions Using Multiphase Medical Images

Algorithm 1

Codebook learning via OMP and K-SVD.

(1) Task: Find the optimal codebook that represent the data

  samples as sparse compositions. Objective function is:

(2) Init: Set dictionary matrix with normalized columns

(3) while (stopping rule not met) do

(4) Sparse Coding Stage: Compute to approximate example

    using OMP, by solving:

(5) Codebook Update Stage: update each dictionary atom ,

(6)  - -Define the group of indices corresponding to atom ,

(7)  - -Compute the overall representation error:

(8)  - -Obtain by restricting using

(9)  - -Apply SVD: . Atom is updated by the first

     column of and coefficient vector is updated by the

     first column of multiplied by

(10)  - -Padding and to their original size by 0 according

      to

(11) end while