Low Illumination Image Enhancement Method Based on Deep Convolutional Neural Networks
The image details and textures under low illumination conditions are difficult to distinguish,resulting in serious information loss.The traditional enhancement method requires a lot of manual adjustment for parameters,low efficiency and no prominent details after enhancement.To solve this problem,a low illumination image enhancement model based on Convolutional Neural Networks(CNN)is proposed.It automatically learns the decomposition and enhancement of low illumination image through a data-driven network structure,and updates model parameters by end-to-end training.This model includes modules of decomposition network,illumination adjustment network and noise reduction,and Convolutional Block Attention Module(CBAM)is added to the decomposition network and the illumination adjustment network,to capture important information in the image more comprehensively.This model firstly decomposes the image into the illumination component and the reflection component by the decomposition network,and then inputs the illumination adjustment network and noise reduction module respectively for processing,and finally reconstructs to obtain the enhanced image.The experimental results demonstrate that compared to other common enhancement algorithms,this method effectively improves the contrast and texture details of low illumination image,providing clearer and more reliable image quality.
国家科技期刊平台
NSTL
万方数据
