ELGC-Net is an efficient change detection framework for remote sensing image pairs, designed to accurately identify semantic changes while reducing model size. The framework includes a Siamese encoder, fusion modules, and a decoder. A key component is the Efficient Local-Global Context Aggregator (ELGCA) module, which captures both local and global contextual information using a novel pooled-transpose (PT) attention and depthwise convolution. The PT attention reduces computational complexity by leveraging pooling operations and transposed attention, while depthwise convolution captures local spatial features. ELGC-Net outperforms existing methods on three challenging datasets, achieving a 1.4% improvement in intersection over union (IoU) on the LEVIR-CD dataset compared to ChangeFormer, with significantly fewer parameters. A lightweight variant, ELGC-Net-LW, further reduces computational complexity while maintaining comparable performance. The framework demonstrates superior performance in both quantitative and qualitative evaluations, effectively detecting subtle and significant changes in remote sensing images. The ELGCA module efficiently aggregates contextual information, reducing parameters and computational cost while maintaining high accuracy. The model is implemented with a Siamese architecture, and experiments show that it achieves state-of-the-art results across multiple datasets. The framework is efficient, scalable, and suitable for resource-constrained environments.ELGC-Net is an efficient change detection framework for remote sensing image pairs, designed to accurately identify semantic changes while reducing model size. The framework includes a Siamese encoder, fusion modules, and a decoder. A key component is the Efficient Local-Global Context Aggregator (ELGCA) module, which captures both local and global contextual information using a novel pooled-transpose (PT) attention and depthwise convolution. The PT attention reduces computational complexity by leveraging pooling operations and transposed attention, while depthwise convolution captures local spatial features. ELGC-Net outperforms existing methods on three challenging datasets, achieving a 1.4% improvement in intersection over union (IoU) on the LEVIR-CD dataset compared to ChangeFormer, with significantly fewer parameters. A lightweight variant, ELGC-Net-LW, further reduces computational complexity while maintaining comparable performance. The framework demonstrates superior performance in both quantitative and qualitative evaluations, effectively detecting subtle and significant changes in remote sensing images. The ELGCA module efficiently aggregates contextual information, reducing parameters and computational cost while maintaining high accuracy. The model is implemented with a Siamese architecture, and experiments show that it achieves state-of-the-art results across multiple datasets. The framework is efficient, scalable, and suitable for resource-constrained environments.