In this study, a low light image enhancement approach using autoencoder-based deep neural networks is proposed. The proposed approach consists of two subnets, namely, feature extraction (FE) subnet and color enhancement (CE) subnet. FE subnet extracts image features from the low light image and enhances detailed textures, while CE subnet recovers the color information of the low light image and performs image denoising. A pre-processing technique, namely, relative global histogram stretching (RGHS) is employed in CE subnet. Finally, the processing results of FE and CE subnets are fused to generate the final enhanced image. Based on the experimental results obtained in this study, in terms of two objective performance metrics (PSNR (dB) and SSIM) and subjective evaluation, the performance of the proposed approach is better than those of five comparison approaches.