Multi-scale Feature Fusion Algorithm Based on Generative Adversarial Network for Motion Deblurring

The engine vibration of the gasoline-powered unmanned helicopter induces motion blur in the images acquired by the onboard camera, degrading the performance of visual tasks such as target identification and visual navigation. To address this issue, this paper proposes a multi-scale feature fusion deblurring algorithm based on a generative adversarial network (GAN). The algorithm utilizes different scales of image features to enhance both semantic and detailed information in the images. Additionally, a frequency selection method is employed to extract useful frequency information from the blurred images, further improving the deblurring effect. Experimental validation is conducted on the publicly available GoPro dataset, where the results demonstrate that the algorithm achieves a PSNR (Peak Signal-to-Noise Ratio) of 32.24dB and an SSIM (Structural Similarity Index) of 0.934. Furthermore, the algorithm is tested on a dataset captured by the gasoline-powered unmanned helicopter, showcasing its effectiveness in enhancing target identification performance for motion-blurred images. This finding provides strong support for the visual task capabilities of gasoline-powered unmanned helicopters.