<p>Low-light image enhancement (LLIE) is an important low-level vision task. Current LLIE methods pay less attention to building their models under color-based image representation framework. In this paper, based on the HVI color space, we propose a simple yet effective dual-branch Mamba-Transformer network (DBMT) for the LLIE task, aiming to obtain high-quality enhancement with satisfying computational efficiency. DBMT has two encoding-decoding branches for learning mappings of image intensity and color between dark/bright image pairs. At each feature resolution, we employ Mamba-Transformer Mixer (MTM) block to extract intensity and color features and further refine them through exploring their mutual associations. With these MTM blocks, DBMT combines the computational efficiency of Mamba with the global context capturing ability of Transformer. We conducted quantitative and qualitative experiments on several public datasets to verify the effectiveness of our method. The results demonstrate the superiority our model over many state-of-the-art LLIE methods.</p>