Representative Color Transform for Image Enhancement
作者:Hanul Kim1, Su-Min Choi2, Chang-Su Kim3, Yeong Jun Koh
單位:Seoul National University of Science and Technology 2Chungnam National University 3Korea University
Abstract
前人方法都是encode-decode方式,丟失細(xì)節(jié);密集轉(zhuǎn)化也限制顏色空間的遷移效果;
本文使用顏色遷移表征(RCT)表征顏色變化,根據(jù)輸入和表征顏色相似性增強(qiáng)顏色,得到更好效果;
RCT determines different representative colors specialized in in- put images and estimates transformed colors for the repre- sentative colors. It then determines enhanced colors us- ing these transformed colors based on the similarity be- tween input and representative colors.
Introduction
- 問題
First, details of the input im- age are not preserved in the up-sampling process of the de- coder, even though they employ skip-connections. Second, these approaches train networks with fixed input size, which makes it difficult to enhance images of arbitrary spatial res- olutions in the inference phase.
1.使用上采樣無法保證解碼器還原細(xì)節(jié)特征;
2.固定尺寸輸入讓輸入圖片有限制;
還有一種方法:全局參數(shù)估計,不需要上采樣,使用RGB, CIELab,LUTS等方式,但不同通道之間無法分開訓(xùn)練顏色遷移;
- 本文使用方法
RCT學(xué)習(xí)大規(guī)模色彩遷移,encode+特征表征=大規(guī)模色彩遷移能力
Related work
- 數(shù)據(jù)集MIT-Adobe 5K,深度學(xué)習(xí)開始的方法直接學(xué)習(xí)像素級,端到端,但效果不行;
- 逐漸出現(xiàn)encoder-decoder方法,從GAN到預(yù)測光流網(wǎng)絡(luò),到頻率分解方法;
- 全局參數(shù)估計:使用密集遷移方程、通道密集遷移、強(qiáng)化學(xué)習(xí)、3D LUTS,預(yù)定義無法有效遷移;
Method
得到模型結(jié)構(gòu),代碼如下
MagicGeorge/RCTNet
實驗
在四個實驗數(shù)據(jù)集上測試,達(dá)到不錯結(jié)果;文章來源:http://www.zghlxwxcb.cn/news/detail-511109.html
Conclusion
使用表征和顏色轉(zhuǎn)換特征,利用全局和局部特征融合,得到對應(yīng)顏色矩陣,提高色彩強(qiáng)化效果。文章來源地址http://www.zghlxwxcb.cn/news/detail-511109.html
到了這里,關(guān)于圖像色彩增強(qiáng)相關(guān)論文閱讀-Representative Color Transform for Image Enhancement(ICCV2021)的文章就介紹完了。如果您還想了解更多內(nèi)容,請在右上角搜索TOY模板網(wǎng)以前的文章或繼續(xù)瀏覽下面的相關(guān)文章,希望大家以后多多支持TOY模板網(wǎng)!