Adaptive CNN-Based Image Compression Model for Improved Remote Desktop Experience

This paper addresses the optimization of desktop image presentation in remote desktop scenarios. Remote desktop tools, essential for work efficiency, often employ image compression to manage bandwidth. While JPEG is a prevalent choice due to its efficiency in eliminating redundancy, it can introduce artifacts as compression increases. Recently, deep learning-based compression techniques have emerged, rivaling traditional methods like JPEG. This research introduces a convolutional neural network-based model for image compression and reconstruction, emphasizing human visual perception. By integrating adaptive spatial and channel attention mechanisms, it ensures better preservation of text and texture. This method outperforms JPEG and other deep learning algorithms in image quality and compression ratio.