Premium
68‐2: Data‐driven Image Enhancement Using Deep Neural Networks for a Display Image Pipeline
Author(s) -
Na Sewhan,
Jang Woohyuk,
Lim Hyunwook,
Lee Jaeyoul,
Kang Imsoo
Publication year - 2019
Publication title -
sid symposium digest of technical papers
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.351
H-Index - 44
eISSN - 2168-0159
pISSN - 0097-966X
DOI - 10.1002/sdtp.13085
Subject(s) - pipeline (software) , artificial intelligence , computer science , artificial neural network , image (mathematics) , pixel , computer vision , transformation (genetics) , deep learning , pattern recognition (psychology) , biochemistry , chemistry , gene , programming language
A data‐driven image reproduction algorithm on the conventional simple image pipeline with deep neural networks is proposed. The deep neural networks determining the pipeline parameter values are learned by end‐to‐end training of the pipeline. The algorithm is efficient because the network operates in low‐resolution mode and the image operators for the pixel‐wise image full‐resolution transfers are ready in most of display image pipeline hardware. A user study on the challenging images indicates that the model performs well on complicated content‐dependent image enhancement transformation: the image appearance is preferred to several other methods.