The paper addresses the challenges in Few-Shot Fine-Grained Image Classification (FS-FGIC) by proposing a novel network called C2-Net. The main issues addressed are the extraction of discriminative features while reducing sample-level noise and achieving accurate feature matching between support and query samples with variable spatial positions and angles. C2-Net consists of two main modules: the Cross-Layer Feature Refinement (CLFR) module and the Cross-Sample Feature Adjustment (CSFA) module. The CLFR module integrates and refines features from multiple layers to suppress noise and non-generalizable information, while the CSFA module adjusts query features to align with support samples through channel activation and position matching operations. Extensive experiments on five fine-grained benchmark datasets demonstrate that C2-Net outperforms state-of-the-art methods, showing superior performance in both fine-grained discrimination and few-shot learning tasks.The paper addresses the challenges in Few-Shot Fine-Grained Image Classification (FS-FGIC) by proposing a novel network called C2-Net. The main issues addressed are the extraction of discriminative features while reducing sample-level noise and achieving accurate feature matching between support and query samples with variable spatial positions and angles. C2-Net consists of two main modules: the Cross-Layer Feature Refinement (CLFR) module and the Cross-Sample Feature Adjustment (CSFA) module. The CLFR module integrates and refines features from multiple layers to suppress noise and non-generalizable information, while the CSFA module adjusts query features to align with support samples through channel activation and position matching operations. Extensive experiments on five fine-grained benchmark datasets demonstrate that C2-Net outperforms state-of-the-art methods, showing superior performance in both fine-grained discrimination and few-shot learning tasks.