The work aims to propose a color constancy-based image enhancement method under poor illumination regarding the low image brightness and contrast caused by such factors as poor illumination or underexposure in the process of color image acquisition. The HSV color space was used to eliminate the correlation between the color components and the hue component remained unchanged to avoid color distortion. On the one hand, the luminance component was enhanced with the improved MSR (Multi-scale Retinex) algorithm to strengthen the brightness and contrast of the image. On the other hand, the saturation component was subject to adaptive nonlinear stretching to improve color saturation. The proposed method could efficiently improve the contrast and information entropy of the image, and achieve better visual effects. Compare with the traditional MSR and MSRCR algorithms, the objective evaluation indicators of the proposed method were better, and the proposed method was much faster. The proposed method can efficiently improve the brightness and contrast of the image under poor illumination, and has a stronger ability to maintain color and reproduce image details. Experimental results demonstrate the effectiveness of the proposed method.