TABLE 2. Convolutional Neural Network-Based Super-Resolution.
Method | Featured structure | Average Computation time |
---|---|---|
Super-Resolution Convolutional Neural Network (SRCNN) | No pooling process | 0.39 s per image in dataset at magnification power of 3 [9] |
Fast Super-Resolution Convolutional Neural Network (FSRCNN) | No pooling process and additional process of shrinking and expanding | 0.061 s per image in dataset at magnification power of 3 [11] |
Deeply-Recursive Convolutional Network (DRCN) | No pooling process, partially recursive layers, and skip connection | Not measured in [12] |