<p>Style transfer is a technique that applies the artistic style of one image to another, with widespread applications in image generation and processing. However, traditional style transfer methods often struggle to balance the accurate representation of style features and the integrity of content structure, especially in real-time applications where processing efficiency is low. To address these issues, we propose a new method that combines multi-scale feature fusion with advanced attention mechanisms. Specifically, we extract multi-scale features from different layers of a deep convolutional network to capture both fine details and overall information. Meanwhile, we employ a multi-head attention mechanism to process different regions of the image in parallel, enhancing feature diversity and precision. Additionally, we introduce a hybrid attention mechanism that integrates channel and spatial attention, utilizing different convolutional kernel sizes to enhance focus on key visual regions. This approach improves style expression while preserving content integrity. Experimental results show that the proposed method achieves high-quality style transfer on standard datasets and significantly enhances processing efficiency, making it suitable for real-time image and video style transfer applications.</p>