A Multiscale Generative Adversarial Network-Based Adaptive Feature Enhancement for Exhibition Imaging
Zilu Cheng, Hao Li · Journal of Circuits Systems and Computers · 2025
With the rapid development of the exhibition industry, high-quality imaging technology plays an increasingly important role in enhancing the audience experience and transmitting exhibition information. However, due to the complex and changeable exhibition environment, traditional imaging methods are often difficult to obtain satisfactory image quality. Therefore, an adaptive feature enhancement method for exhibition imaging based on a Multi-scale Generative Adversarial Network (MS-GAN) is proposed in this paper. By constructing an MS-GAN model, this method realizes feature extraction and fusion of exhibition images at different scales. The generator part adopts a multi-scale convolution structure, which can capture the local details and global structure information of the image and improve the quality of the generated image through residual connection and attention mechanism. In the discriminator part, the multi-scale feature discriminant strategy is used to evaluate the generated image to guide the optimization direction of the generator. The experimental results show that this method can significantly improve the brightness, clarity and color saturation of exhibition images and effectively improve the image quality. Especially under complex lighting conditions and noise interference, the method can still maintain good performance stability and enhancement effect. In addition, the method also has strong adaptive ability, which can automatically adjust the enhancement parameters according to different exhibition environments and imaging conditions to achieve a personalized image enhancement effect.