The classification of colorectal cancer histopathological images plays a critical role in the early diagnosis of colorectal cancer. However, existing classification methods for colorectal histopathological images have not fully considered the highly irregular nature and structural heterogeneity of colorectal cancer glands, making it difficult to effectively fuse global and local features. This limitation has led to suboptimal performance in the classification of colorectal cancer histopathological images. To enhance the accuracy of such classification, this paper proposes a Colorectal Cancer Histopathological Image Classification Method Based on Complementary Fusion of Global and Local Features (CFGL). First, in CFGL, the Multi-Region Lesion Feature Aggregation (MRFA) module and Residual KAN (RKAN) module are designed to enhance the capability of capturing global and local features of glands, respectively. Then, within the Glandular Lesion Region Enhancement (GLRE) module, the Multi-scale Laplace Feature Extraction (MLFE) module is developed to extract gland features at different scales, effectively leveraging high-frequency edge information to refine feature extraction. Finally, the Dynamic Interactive Matrix Feature Fusion (DIMF) module is proposed to perform complementary fusion of global and local features by further enhancing the contextual awareness and sensitivity to gland regions. Experimental results on the two datasets of MHIST and MedMfc_Colon show that CFGL has better classification performance than existing methods.