Super-Resolution Deblurring Algorithm for Generative Adversarial Networks
Tian Bing, Wentao Yan, Wei Wang, Qi Su, Yin Liu, Guangxiu Liu, Wanguo Wang · 2017
Image quality improvement has a significant impact on target detection and recognition. Generative Adversarial Nets are inspired by two-person zero-sum game in game theory. It can learn to automatically generate images, which can be conditional learning, a guide to the image generation. In this paper, we analyze the characteristics of motion blur, and propose a method to add the defocused fuzzy kernel and multi-direction motion fuzzy kernel to the training samples, and use the super-resolution anti-network method to exercise blur and carry out the fuzzy image data recorded by the UAV experiment analysis.