GAN (Generative Adversarial Nets)-based CFA (Color Filer Array) image demosaicing joint denoising method

Abstract

The invention discloses a GAN (Generative Adversarial Nets)-based CFA (Color Filer Array) image demosaicing joint denoising method. The method comprises the following steps: (1) a training sample setis acquired; (2) the GAN is built; (3) parameters of a 9-layer convolution neural network are updated; (4) parameters of a 39-layer convolution neural network are updated; (5) whether the updating times of the 39-layer convolution neural network and the 9-layer convolution neural network reach 200 times is judged, if yes, a step (6) is executed, or otherwise, the step (3) is executed; (6) a nonlinear mapping relationship is built; and (7) an image after demosaicing and denoising is acquired. The color information of a noisy CFA image obtained by a digital camera can be well recovered, the noise introduced during an image acquisition process of the digital camera can be effectively suppressed, the appearance of unnatural colors is reduced, and the visual effects of a color image are improved.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a color filter array CFA (ColorFiler Array) image demosaic joint denoising method based on Generative Adversarial Nets (GAN) in the technical field of digital camera image restoration. The present invention can be used to restore the complete color information of the noise-containing color filter array CFA image in a single-sensing camera, thereby solving the problems of loss of image color information and introduction of noise caused by reducing the cost of camera hardware; at the same time, it can also be used for The original data saved by the camera in "Raw Mode" is processed in the computer to obtain high-quality color images. Background technique [0002] Single-sensing cameras use a single CCD (Charge Coupled Deyice) or CMOS (Complementary Metal Oxide Semiconductor) sensor chip to capture images by covering a color filter array CFA in front of the single sen...

AI Technical Summary

Problems solved by technology

The shortcomings of this method are: there are many areas containing unnatural colors, and the noise removal is not clean. More importantly, the noise model learned by this method from a large amount of training data is tailored for a single noise level. Therefore, the algorithm is less robust to noise variance

However, the shortcomings of this method are: the shallow convolutional neural network is not comprehensive enough to learn the prior information of the color image, and cannot fully mine the more detailed structural information in the image, resulting in artifacts in the restored image. , the loss of details such as edges is relatively serious, and the overall image tends to be smooth, which eventually leads to unsatisfactory visual effects

Images