Reconstructed Error and Linear Representation Coefficients Restricted by -Minimization for Face Recognition under Different Illumination and Occlusion

Algorithm 1

The proposed method.

Input: The training samples for classes,

denotes the training sample matrix from th class, denotes the

number of training images of the th individuals. A test sample .

   .

   ,

   Let .

   Compute:

   for each subject   do

    (a) Use formula (13)–(29) to solve the objective function

          ,

          ,

    (b) the solution is

          ,

    (c) the unknown variables are

   end for

   Calculate the -norm of ,

   Calculate the Gradient of .

Output: the recognition protocols:

Reconstruction error based method,