Infrared Small Target Detection (IRSTD) plays a vital role in military surveillance and national defense. However, it remains a highly challenging task due to the small size of targets, low target-to-background contrast, and complex background clutter. To tackle these issues, we propose a novel detection framework, the Dual-branch Contrast-Attentive Pyramid Network (DCAPNet), which is designed to model the contrastive relationships between targets and background across multiple feature levels to improve detection performance.Specifically, the Multi-Contrast Dilation (MCD) module enhances weak target responses by leveraging multi-scale dilated convolutions to capture essential semantic cues. The Dual Contrast Enhance (DCE) module adopts a dual-branch architecture to emphasize local details and alleviate representation bottlenecks in low-texture regions. Moreover, the Contrast-Attention Fusion (CAF) module facilitates multi-scale semantic alignment and introduces an attention-guided mechanism to enhance edge perception while suppressing background interference. Extensive experiments on three publicly available IRSTD benchmarks, NUAA-SIRST, IRSTD-1K, and NUDT-SIRST, demonstrate that DCAPNet consistently outperforms state-of-the-art methods across various key evaluation metrics. These results validate the effectiveness, robustness, and generalization capability of the proposed method in complex infrared imaging scenarios.