Weak light image enhancement deep learning algorithm based on channel attention

The invention relates to the technical field of weak light image enhancement, in particular to a weak light image enhancement deep learning algorithm based on channel attention, which comprises the following steps of: firstly, sending an input weak light image into a pixel illumination estimation mo...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: ZHANG ZHENGQING, HU CHAO, YU MENG, WU WEIJIE, HUANG JIAYAO, ZHU LIQIANG, LAI SHENGXIN
Format: Patent
Sprache:chi ; eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The invention relates to the technical field of weak light image enhancement, in particular to a weak light image enhancement deep learning algorithm based on channel attention, which comprises the following steps of: firstly, sending an input weak light image into a pixel illumination estimation module for feature extraction, and adjusting an illumination adjustment coefficient required by each pixel through gain estimation; obtaining a gain coefficient of each pixel, and multiplying the extracted features by the gain coefficients to obtain a preliminary illumination image; the obtained preliminary illumination image is sent to a channel attention block, information such as structures and modes in important channels is enhanced, meanwhile, denoising processing is carried out, a result of the channel attention block is input to a fine tuning module, fusion fine tuning is carried out on features of different receptive fields and original features, and final features are obtained; and carrying out operations su