Image processing apparatus, image processing method, and non-transitory computer-readable storage medium

An image processing method comprising:

• acquiring a set of patches that includes a patch of interest formed of a plurality of pixels relative to a pixel of interest and a similar patch that is similar to the patch of interest in an input image;
• estimating a noise amount parameter that corresponds to each pixel in the patch of interest based on a pixel value of the each pixel in the patch of interest;
• calculating, based on the set of patches, a covariance matrix that corresponds to the set of patches on the assumption that each pixel has a uniform noise amount parameter;
• estimating, based on the covariance matrix and the noise amount parameter, a covariance matrix that corresponds to the noise amount parameter of the each pixel;
• and
• executing, using the estimated covariance matrix, a process of generating a first patch in which noise of at least the patch of interest among the patches included in the set of patches is reduced.

A non-transitory computer-readable storage medium storing a computer program for causing a computer to function as:

• an acquiring unit that acquires a set of patches that includes a patch of interest formed of a plurality of pixels relative to a pixel of interest and a similar patch that is similar to the patch of interest in an input image;
• a first estimation unit that estimates a noise amount parameter that corresponds to each pixel in the patch of interest based on a pixel value of each pixel in the patch of interest;
• a first calculation unit that calculates, based on the set of patches, a covariance matrix that corresponds to the set of patches on the assumption that each pixel has a uniform noise amount parameter;
• a second estimation unit that estimates, based on the covariance matrix and the noise amount parameter, a covariance matrix that corresponds to the noise amount parameter of each pixel;
• and
• a processing unit that executes, using the covariance matrix estimated by the second estimation unit, a process of generating a first patch in which noise of at least the patch of interest among the patches included in the set of patches is reduced.