Method for upscaling noisy images, and apparatus for upscaling noisy images
With known image upscaling methods of noisy images, important detail information is lost during denoising. A method for upscaling noisy input images comprises upscaling a noisy input image to obtain a noisy High-Resolution (HR) image, denoising the noisy input image to obtained a denoised Low-Resolution (LR) image, upscaling the denoised LR image to obtain an upscaled denoised LR image, and combining the noisy HR image and the upscaled denoised LR image to generate a denoised HR image.
1. A method for upscaling noisy input images, comprising:
upscaling a noisy input image to obtain a noisy High-Resolution, HR, image;
denoising the noisy input image to obtain a denoised Low-Resolution, LR, image;
upscaling the denoised LR image to obtain an upscaled denoised LR image; and
combining, by a series of non-linear filtering steps, the noisy HR image and the upscaled denoised LR image to generate a denoised HR image.
2. The method according to claim 1 , wherein for said upscaling the noisy input image and for said upscaling the denoised LR image the same upscaling method is used.
3. The method according to claim 1 , wherein the non-linearity is obtained through thresholding.
4. The method according to claim 1 , wherein the denoising uses a five stage filtering sequence of first denoising filtering, first thresholding, second denoising filtering, second thresholding and third denoising filtering.
5. The method according to claim 1 , wherein the upscaling uses a five stage filtering sequence of first upscaling filtering, first thresholding, second upscaling filtering, second thresholding and third upscaling filtering.
6. The method according to claim 1 , wherein the combining uses a five stage filtering sequence of first combining filtering, first thresholding, second combining filtering, second thresholding and third combining filtering.
7. The method according to claim 3 , wherein the thresholding sets negative values to zero and lets positive values pass unchanged.
8. The method according to claim 4 , wherein coefficients and dimensions of filters for the filtering are initialized randomly and then trained in a training stage by using a machine learning algorithm.
9. The method according to claim 3 , wherein dimensions of filters for the filtering are predefined, and coefficients of filters for the filtering are initialized randomly and then trained in a training stage by using a machine learning algorithm.
10. The method according to claim 1 , wherein said upscaling, denoising and combining use image patches.
11. An apparatus for upscaling noisy input images, comprising
a first image upscaler adapted for upscaling a noisy input image to obtain a noisy High-Resolution, HR image;
an image denoiser adapted for denoising the noisy input image to obtain a denoised Low-Resolution, LR, image;
a second image upscaler adapted for upscaling the denoised LR image to obtain an upscaled denoised LR image; and
an image combiner adapted for combining pixels or patches of the noisy HR image and pixels or patches of the upscaled denoised LR image to generate a denoised HR image, wherein the image combiner is configured to use a network of non-linear filters.
12. The apparatus of claim 11 , wherein for said upscaling the noisy input image and for said upscaling the denoised LR image the same upscaling method is used.
13. The apparatus of claim 11 , wherein the non-linearity is obtained through thresholding.
14. The apparatus according to claim 11 , wherein the image denoiser is configured to use a five stage filtering sequence of first denoising filtering, first thresholding, second denoising filtering, second thresholding and third denoising filtering.
15. The apparatus according to claim 11 , wherein the upscaling uses a five stage filtering sequence of first upscaling filtering, first thresholding, second upscaling filtering, second thresholding and third upscaling filtering.
16. A non-transitory computer-readable storage medium comprising computer-executable instructions to enable a computer to perform a method for upscaling noisy input images, comprising:
upscaling a noisy input image to obtain a noisy High-Resolution, HR, image;
denoising the noisy input image to obtain a denoised Low-Resolution, LR, image;
upscaling the denoised LR image to obtain an upscaled denoised LR image; and
combining, by a series of non-linear filtering steps, the noisy HR image and the upscaled denoised LR image to generated a denoised HR image.