Chinese traditional ethnic patterns usually have unique artistic style expressions. However, existing style transfer methods often suffer from monotonous styles and fragmentation between content patterns and background styles when innovating on these patterns. To address these issues, we propose a novel style transfer model named DFI-EPST, which leverages Dual-Feature Integration to enhance the harmony between content patterns and background styles, resulting in more diverse and natural stylized images. Specifically, DFI integrates two key feature manipulation techniques: background feature embedding and feature transformation. The background feature embedding module uses VGG to extract the underlying pixel-level features of the style image and embeds them into the blank background of the content image, ensuring that style features are naturally integrated into the background. This enhances overall visual harmony and style diversity. Then the feature transformation module employs Adaptive Instance Normalization to, adjusting the mean and variance of content image features to match those of the style image, achieving efficient style transfer. Additionally, DFI-EPST includes a feature shrinkage module that uses a soft thresholding mechanism to retain key content features, ensuring that important details are preserved. The experimental results show that our model effectively solves the problems existing in the traditional style transfer methods, and significantly improves the metrics of CF, SSIM and transfer speed.