Abstract
In this paper, we propose a content adaptive demosaicing algorithm, utilising content analysis and correlation between the red, green and blue planes of a particular image. These two aspects are used for the classification of the technique in the generated trained filters. The proposed method aims to reconstruct a high quality demosaiced image from a CFA Bayer pattern. The strategy highlighted in this paper is very effective, as many of the image details are maintained during reconstruction. Since image content analysis and filter coefficient optimisation are performed during training and the training process is offline, the online de-mosaicing filter is very efficient.
Original language | English |
---|---|
Pages (from-to) | 222-228 |
Journal | Neurocomputing |
Volume | 83 |
Early online date | 13 Jan 2012 |
DOIs | |
Publication status | Published - 15 Apr 2012 |
Keywords
- Bayer pattern
- Demosaic
- CFA array
- RGB
- Classification
- Trained filter