Few-shot semantic segmentation (FSS) suffers significant performance degradation under sample-scarce conditions due to increased intra-class variance and aggravated inter-class confusion, leading to a pronounced shift in feature distribution. Existing methods predominantly focus on foreground objects, neglecting the use of background information and cross-category common semantic priors. This study proposes a novel FSS framework based on common semantic prior, which corrects feature distributions through cross-category shared knowledge, enabling efficient segmentation. On PASCAL-5ⁱ and COCO-20ⁱ under 1-shot setting, the proposed method achieves mIoU of 68.5% and 48.3%, respectively, with only 2.5M parameters, significantly outperforming current state-of-the-art (SOTA) methods. By explicitly modeling cross-category common attributes, the proposed approach effectively mitigates data distribution bias, providing a new paradigm for lightweight FSS.