研究目的
To improve the quality of images captured by sensors in low illumination conditions by enhancing brightness and contrast while avoiding color distortion and over-enhancement.
研究成果
The proposed algorithm effectively improves the brightness and contrast of low-light sensor images while avoiding color distortion and over-enhancement. It outperforms other state-of-the-art methods in terms of PSNR, MSE, and LOE, and is more realistic and natural with less distortion. Future work will focus on optimizing the network model to further improve the brightness of dark areas.
研究不足
The algorithm may not enhance dark regions of a large white area in the actual scene as effectively as other regions. Additionally, the brightness improvement in particularly dark areas may not be as good as with the LIME algorithm.
1：Experimental Design and Method Selection:
The study proposes a low-light sensor image enhancement algorithm based on the HSI color model, utilizing a dataset generation method based on the Retinex model, segmentation exponential method for saturation enhancement, and a Deep Convolutional Neural Network for intensity enhancement.
2：Sample Selection and Data Sources:
500 images from the BSD dataset were used as reflective components, and 256,000 image patches of size 40 × 40 were randomly selected for training.
3：List of Experimental Equipment and Materials:
A PC with 16G RAM, an Intel (R) Core (TM) i7-7700 CPU
4：30GHz, and an Nvidia GeForce GTX1070 GPU were used. Experimental Procedures and Operational Workflow:
The low-light image is converted from RGB to HSI color space, the saturation component is enhanced using the segmentation exponential method, and the intensity component is enhanced using a specially designed DCNN. The enhanced image is then transformed back to RGB space.
5：Data Analysis Methods:
Performance was evaluated using peak signal-to-noise ratio (PSNR), structural similarity (SSIM), mean square error (MSE), and lightness order error (LOE).
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
