In order to improve the visibility of low-illumination remote sensing images, an improved multiscale Retinex combined with a local contrast adaptive adjustment method was proposed. First, the original image was transformed into HSI color space and the hue component H, saturation component S, and brightness component I were effectively separated. The H component was unchanged, and an improved multiscale Retinex algorithm was applied to process the I component, to improve the overall brightness and contrast of the image. In this case, the Sigmoid function was used to replace the logarithm function in the multiscale Retinex algorithm to reduce the loss of image data. In order to improve the local detail information, local contrast adaptive enhancement was performed via image processing. Then the component S was processed by piecewise linear enhancement. Finally, the processed image was transformed to RGB color space. The experimental results indicate that the entropy of the image information is increased from 5.79 to 6.65, and the local contrast of the image interest area increased from 0.695 to 0.701. This indicates that the image quality and the applied value were effectively improved.