| CPC G06T 5/70 (2024.01) [G06T 5/40 (2013.01); G06T 5/50 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] | 18 Claims |
|
1. An apparatus comprising:
at least one processor; and
a memory coupled to the at least one processor, the memory having instructions that, when executed by the processor, causes the at least one processor to function as:
an acquisition unit configured to acquire an image group including a plurality of first images;
an adjustment unit configured to adjust luminance range of each of the plurality of first images by applying a gain to each pixel as to narrow the luminance range in order to add noise;
a generation unit configured to generate a plurality of second images by adding the noise to each of the plurality of first images and applying a digital gain, the noise reproducing a histogram containing discrete distribution decreases, and generate learning data including respective pairs of the plurality of first images and the corresponding plurality of second images; and
a learning unit configured to perform learning using a neural network and the learning data,
wherein the generation unit includes:
a first adjustment unit configured to generate a plurality of luminance adjusted images by adjusting brightness of each of the plurality of first images,
an addition unit configured to generate a plurality of noise added images by adding the noise to each of the plurality of first images or each of the plurality of luminance adjusted images,
a conversion unit configured to perform integer conversion of each of the plurality of noise added images, and
a second adjustment unit configured to generate the plurality of second images by adjusting brightness of each of the plurality of noise added images subjected to the integer conversion,
wherein the learning unit includes:
a first learning unit configured to perform first learning using respective pairs of the plurality of first images and a corresponding plurality of second images that is generated by adjusting the brightness of each of the plurality of first images and adding the noise to each of the plurality of first images, as learning data, and
a second learning unit configured to perform second learning using respective pairs of the plurality of first images and a corresponding plurality of second images that is generated by adding the noise to each of the plurality of first images without adjusting the brightness of each of the plurality of first images, as learning data, and
wherein the first learning unit performs the first learning using the learning data including the generated plurality of second images.
|