Space Target Super-resolution Based on Low-complex Convolutional Networks

The images super-resolution is a key technique for space target, which has important research significance and application value. However, the traditional super-resolution algorithms always fail to achieve a good reconstruction effect due to limited information and inaccurate manual design models. Moreover, deep learning algorithms mainly focus on improving reconstruction with the price of high complexity. To address the above issues, we propose an improved space target super-resolution based on low-complex convolutional neural networks. Specifically, we construct a three-layer convolutional neural network by changing network layers and convolution kernels, which reduces the number of parameters and improves the extraction characteristics and reconstruction. Meanwhile, we establish a complete training data set, which solves the problem that the space target images are difficult to obtain. The experimental results demonstrate that the algorithm proposed can achieve the space target super-resolution. Compared with other algorithms, our algorithm obtains better subjective visual effects and objective evaluation indicators.