Optimal Superpixel Kernel-Based Kernel Low-Rank and Sparsity Representation for Brain Tumour Segmentation

Algorithm 1

Optimal superpixel kernel generation based on MKL.

Input	: Training sample set and corresponding label set; kernel scale range .
Output	: Optimal superpixel kernel.
	Step 1: Select M kernel scales .
	Step 2: Compute the Gram matrixes with using equation (3).
	Step 3: Vectorize and construct the matrix .
	Step 4: Compute the optimal weight vector by solving equation (7).
	Step 5: Compute the optimal kernel function using equation (8).
	Step 6: Compute the optimal superpixel kernel using equation (9).