Considering the difficulty in obtaining real paired training data for low-light image enhancement and existing methods rarely consider both upstream and downstream visual tasks when developing algorithms, a progressive unsupervised image enhancement method focusing on value aware was designed, which employed a joint loss function with multiple constraints to get rid of paired datasets during the training stage. With the help of the proposed value aware parameter estimation network, the feature extraction could be achieved with only 0.035 M parameters. To enhance the nonlinear adjustment capability and reduce the number of iterations, a high-order nonlinear mapping curve was designed. To verify the effectiveness of the proposed method, the qualitative and quantitative experiments were conducted on an authoritative dataset widely used in the field of image enhancement, indicating that the proposed method outperformed the existing solutions. In addition, the UAV nighttime object tracking was evaluated as a typical downstream visual perception task. The results on the authoritative evaluation benchmark demonstrate the proposed method can significantly improve the performance of the existing tracker in the night scene, with accuracy and success rate gains of 21.50% and 32.23%, respectively. Extensive experiments show that the proposed method can significantly improve the visual effects of low-light images while also effectively alleviate the performance degradation of downstream vision algorithms caused by insufficient illumination in night scenarios.