Di Zhao, Lan Ma, Zhixian Lin, Tailiang Guo, Shanling Lin
{"title":"End-to-End Denoising of Dark Image Using Residual Dense Network","authors":"Di Zhao, Lan Ma, Zhixian Lin, Tailiang Guo, Shanling Lin","doi":"10.1109/ISNE.2019.8896666","DOIUrl":null,"url":null,"abstract":"When taking pictures in extremely dark light, the image taken is usually very dark and noisy due to the small amount of light entering, and there is clear distortion of color. Since the method of using multi-frame or long-exposure has limitations, we use the single-frame denoising method based on residual dense network (RDN), which uses residual dense block (RDB) to makes full use of local features. Our method has achieved better results than state-of-the-art methods. In addition, we have applied the trained model without fine-tuning on photos captured by different cameras and have obtained similar end-to-end enhancements.","PeriodicalId":405565,"journal":{"name":"2019 8th International Symposium on Next Generation Electronics (ISNE)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2019 8th International Symposium on Next Generation Electronics (ISNE)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ISNE.2019.8896666","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
When taking pictures in extremely dark light, the image taken is usually very dark and noisy due to the small amount of light entering, and there is clear distortion of color. Since the method of using multi-frame or long-exposure has limitations, we use the single-frame denoising method based on residual dense network (RDN), which uses residual dense block (RDB) to makes full use of local features. Our method has achieved better results than state-of-the-art methods. In addition, we have applied the trained model without fine-tuning on photos captured by different cameras and have obtained similar end-to-end enhancements.