Image processing apparatus, image capture apparatus, and image processing method
An image processing apparatus for controlling white balance of an image is provided. The apparatus detects a white region from image data and calculates a first white balance correction value based on data of the white region. The apparatus also detects, with use of machine learning, a region of a subject that has a preset specific color, from the image data, and calculates a second white balance correction value based on a color of the region of the subject. The apparatus calculates a white balance correction value to be applied to the image data, based on the first white balance correction value and the second white balance correction value.
1. An image processing apparatus, comprising:
an image processing circuit that functions as:
a white detection unit that detects a white region from image data;
a first calculation unit that calculates a first white balance correction value based on data of the white region;
a generation unit that generates a detection map representing a detection result of a region of a subject that has a preset specific color, based on information output using machine learning, wherein the detection map represents, for each of partial regions of the image data, reliability based on the detection result;
a second calculation unit that calculates a second white balance correction value based on a color of partial regions determined based on the detection map; and
a third calculation unit that calculates a third white balance correction value based on the first white balance correction value and the second white balance correction value, the third white balance correction value being a white balance correction value that is applied to the image data.
2. The image processing apparatus according to claim 1 , wherein
when reliability of the first white balance correction value is determined to be high, the first white balance correction value is used as the white balance correction value that is applied to the image data, and
when the reliability of the first white balance correction value is not determined to be high, the third white balance correction value is used as the white balance correction value that is applied to the image data.
3. The image processing apparatus according to claim 1 , wherein
the preset specific color is a greenish color or a skin color.
4. The image processing apparatus according to claim 1 , wherein
the third calculation unit calculates the third white balance correction value by performing weighted addition of the first white balance correction value and the second white balance correction value.
5. The image processing apparatus according to claim 4 , wherein
a weight used in the weighted addition is based on reliability of at least one of the first white balance correction value and the second white balance correction value.
6. The image processing apparatus according to claim 5 , wherein
the weight used in the weighted addition is further based on reliability of a condition in which a shooting scene is a specific scene.
7. The image processing apparatus according to claim 6 , wherein
the specific scene is an outdoor scene.
8. The image processing apparatus according to claim 1 , wherein
the second calculation unit calculates the second white balance correction value based on the specific color and on a color temperature of a light source.
9. The image processing apparatus according to claim 8 , wherein
the second calculation unit estimates the color temperature of the light source based on data that has a color included in a preset range among the image data.
10. An image capture apparatus, comprising:
an image sensor;
an image processing apparatus that calculates a white balance correction value from image data obtained by the image sensor, wherein the image processing apparatus comprises an image processing circuit that functions as:
a white detection unit that detects a white region from image data;
a first calculation unit that calculates a first white balance correction value based on data of the white region;
a generation unit that generates a detection map representing a detection result of a region of a subject that has a preset specific color, based on information output using machine learning, wherein the detection map represents, for each of partial regions of the image data, reliability based on the detection result;
a second calculation unit that calculates a second white balance correction value based on a color of partial regions determined based on the detection map; and
a third calculation unit that calculates a third white balance correction value based on the first white balance correction value and the second white balance correction value; and
a development unit that applies development processing to the image data with use of the third white balance correction value.
11. An image processing method executed by an image processing apparatus, the image processing method comprising:
detecting a white region from image data;
calculating a first white balance correction value based on data of the white region;
generating, based on information output using machine learning, a detection map representing a detection result of a region of a subject that has a preset specific color, wherein the detection map represents, for each of partial regions of the image data, reliability based on the detection result;
calculating a second white balance correction value based on a color of partial regions determined based on the detection map; and
calculating a third white balance correction value based on the first white balance correction value and the second white balance correction value, the third white balance correction value being a white balance correction value that is applied to the image data.
12. A non-transitory machine-readable medium that stores a program for causing a computer to execute an image processing method comprising:
detecting a white region from image data;
calculating a first white balance correction value based on data of the white region;
generating, based on information output using machine learning, a detection map representing a detection result of a region of a subject that has a preset specific color, wherein the detection map represents, for each of partial regions of the image data, reliability based on the detection result;
calculating a second white balance correction value based on a color of partial regions determined based on the detection map; and
calculating a third white balance correction value based on the first white balance correction value and the second white balance correction value, the third white balance correction value being a white balance correction value that is applied to the image data.
13. The image processing apparatus according to claim 1 , wherein the information output using machine learning represents a probability that a pixel or block of the image data is the region of the subject that has the preset specific color.
14. The image processing apparatus according to claim 1 , wherein the machine learning is used to output a pixel or block of the image data that has a green color of a plant.