A non-transitory computer-readable storage medium storing a computer program which, when executed by a processor, implements the following:
- determining a target noise parameter of a to-be-denoised image according to an image noise calibration result obtained by pre-performing an image noise calibration on an image acquisition device of the to-be-denoised image;
- performing a preliminary filtering process on the to-be-denoised image to obtain a preliminary filtered image of the to-be-denoised image;
- determining a noise level estimation result of the to-be-denoised image according to the target noise parameter and the preliminary filtered image;
- and
- performing a final denoising process on the to-be-denoised image according to the noise level estimation result to obtain a final blind denoising result of the to-be-denoised image;
- wherein when executed by the processor, the computer program implements pre-performing the image noise calibration on the image acquisition device of the to-be-denoised image by:
- determining at least one candidate exposure gain value according to a device attribute of the image acquisition device;
- for each of the at least one candidate exposure gain value, acquiring data of at least two to-be-calibrated images from a same shooting scene through the image acquisition device, wherein at least one candidate exposure gain value is provided;
- and
- for each of the at least one candidate exposure gain value, determining a candidate noise parameter of a candidate exposure gain value according to the data of the at least two to-be-calibrated images associated with the candidate exposure gain value, and determining the image noise calibration result of the image acquisition device according to each of the at least one candidate exposure gain value and the candidate noise parameter associated with each of the at least one candidate exposure gain value;
- wherein when executed by the processor, the computer program implements, for each of the at least one candidate exposure gain value, determining the candidate noise parameter of the candidate exposure gain value according to the data of the at least two to-be-calibrated images associated with the candidate exposure gain value by:
- determining an image noise level expression of the image acquisition device as follows:
- V
-
- (
- x
- )
- =
- ∑
- i
- =
- 0
- n
- a
- i
-
- x
- i
- ,
- wherein x denotes a target pixel luminance value of a target pixel point, V(x) denotes a noise variance for characterizing a noise level of the target pixel point, n denotes a polynomial series, and aj is a to-be-calibrated noise parameter for characterizing a to-be-calibrated i-th order noise parameter under each of the at least one candidate exposure gain value, wherein n is a positive integer, n≥1, and 0≤i≤n;
- determining an average calibrated image according to a pixel average value of the data of the at least two to-be-calibrated images associated with each of the at least one candidate exposure gain value on each pixel point;
- determining a variance calibrated image according to a pixel value variance of the data of the at least two to-be-calibrated images associated with each of the at least one candidate exposure gain value on each pixel point;
- and
- determining at least (n+1) pairs of pixels with different values from the average calibrated image and the variance calibrated image, determining, according to the at least (n+1) pairs of pixels, a value of the to-be-calibrated noise parameter based on the image noise level expression, and using the value of the to-be-calibrated noise parameter as the candidate noise parameter for each of the at least one candidate exposure gain value.