Joint supervised underwater image enhancement algorithm based on detail enhancement and multi-color space learning
Due to the special underwater imaging environment,serious color offset and atomization usually occurs to underwater images.In view of this,on the basis of the underwater optical imaging model,this paper designs a new enhancement algorithm,named unsupervised underwater image enhancement algorithm based on detail enhancement and multi-color space learning(UUIE-DEMCSL).In this algorithm,an enhancement network based on multi-color space is designed.This network converts the input into multiple color spaces(HSV,RGB and LAB)for feature extraction.The extracted features are fused,so that the network can learn more image feature information and enhance the input image more accurately.The UUIE-DEMCSL is designed according to the underwater optical imaging model and the joint supervised learning framework,which is more suitable for the application scenarios of underwater image enhancement tasks.A large number of experimental results on different data sets show that the proposed UUIE-DEMCSL algorithm can generate underwater enhanced images with good visual quality,and each index of the algorithm has significant advantages.
underwater image enhancementmulti-color space learningunsupervised learningdetail enhancementfeature extractionfeature fusion
万方数据
维普
