A Network with Composite Loss and Parameter-free Chunking Fusion Block for Super-Resolution MR Image

Algorithm 1

Training framework based on the proposed method.

	Input: High-resolution image dataset , magnification times and patch size .
	Output: Super-resolution (SR) Model .
(1)	Initializing the SR model ;
(2)	for in do
(3)	;
(4)	;
(5)	;
(6)	;
(7)
(8)
(9)
(10)
(11)	Back propagation update according to gradient .
(12)	return Model ;