Transform-based image joint demosaicking and denoising method
By using a Transformer-based U-shaped network model, the problems of error accumulation and noise estimation dependence in mosaic images are solved, achieving efficient blind joint demosaicing and denoising, thus improving image quality.
CN116912092BActive Publication Date: 2026-07-21JILIN UNIVERSITY
View PDF 2 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JILIN UNIVERSITY
- Filing Date
- 2023-07-25
- Publication Date
- 2026-07-21
Smart Images
Figure CN116912092B_ABST
Abstract
The present application is suitable for the technical field of image processing, and provides a method for image joint demosaicking and denoising based on a Transformer, comprising the following steps: constructing a data set and pre-processing the data set; adopting a Transformer to build a network model; training the constructed model; and testing and evaluating the trained model, which can restore a clean and noise-free color RGB image from a noisy and mosaicked image, and also solves the problem of dependence of an existing non-blind model on noise estimation accuracy. In addition, an additional module is provided, which can adaptively extract features from the input of the model, thereby helping the model achieve better results.
Need to check novelty before this filing date? Find Prior Art