Single Image Super-Resolution via Cascaded Parallel Multisize Receptive Field

Recovering a High-Resolution (HR) image from a Low-Resolution (LR) image is the main concept of image Super-Resolution(SR). These days Convolution Neural Networks(CNN) have become very popular and efficient in generating HR image from a LR image. Although CNNs are widely used with great performance improvements, there is still much room for improvement. There has always been a trade-off between the number of parameters and performance enhancement. Inspired by the Inception architecture of GoogleNet, we have developed a novel approach with an efficiently increased number of receptive field by cascading different kernel sizes to reconstruct the HR image. Experimental results shows that the proposed model outperforms the state-of-art methods.