Kernel-Based Multiview Joint Sparse Coding for Image Annotation

Algorithm 2

Label prediction procedure.

Input:

  The feature view of test image ,

  Feature view of training images , ,

  Learned dictionary,

  Tuning parameter .

Output:

  Predicted labels for the test image.

(1) Calculate   by Eq. (18)

(2) For  

(3)  Find the five largest values in  and corresponding indices

(4)  Find sample label column vectors in  corresponding to these indices

(5) End For

(6) Compute and rank the weighted frequency of labels appeared in all the found samples, with weights equaling to

the product coefficients of

(7) Transfer labels according to their calculated frequency.