Computing device for instance-based image quality processing and operating method therefor
A computing device for performing image quality processing of an image, and an operating method thereof are provided. The computing device includes: a memory storing at least one instruction; and a processor configured to execute the at least one instruction stored in the memory, to: extract features of an input image by using a first neural network, recognize at least one instance in the input image from the features of the input image by using a second neural network, obtain an image instance quality score map by predicting a quality score corresponding to each instance of the at least one instance in the input image based on the features of the input image by using a third neural network, and perform image quality processing differently for each instance of the at least one instance in the input image by using the quality score corresponding to each instance of the at least one instance in the input image.
1. A computing device comprising:
a memory storing at least one instruction; and
a processor configured to execute the at least one instruction stored in the memory, to:
extract features of an input image by using a first neural network,
recognize a plurality of instances in the input image from the features of the input image by using a second neural network,
obtain an image instance quality score map by predicting a quality score corresponding to each instance of the plurality of instances in the input image based on the features of the input image by using a third neural network, wherein the quality score corresponding to each instance varies depending on whether the instance is in motion or not in motion,
perform image quality processing differently for each instance of the plurality of instances in the input image by using the quality score corresponding to each instance of the plurality of instances in the input image by generating image quality control parameters for each instance in the input image by using the quality score corresponding to each instance in the input image, and performing the image quality processing independently for each instance in the input image by using the image quality control parameters corresponding to the instance, wherein as a value of the quality score corresponding to each instance decreases, an intensity of the image quality processing is applied more strongly, and as a value of the quality score corresponding to each instance increases, the intensity of the image quality processing is applied more lightly, and
generate an output image by merging the image quality-processed instances.
2. The computing device of claim 1 , wherein the image instance quality score map includes position information and the quality score of each instance of the plurality of instances in the input image.
3. The computing device of claim 1 , wherein the first neural network includes at least one convolutional layer.
4. The computing device of claim 1 , wherein the processor is further configured to execute the at least one instruction to obtain an instance map including a class corresponding to each instance, position information of each instance, and a mask indicating a pixel-unit region of each instance, by recognizing the plurality of instances in the input image from the features of the input image by using the second neural network.
5. The computing device of claim 1 , wherein the processor is further configured to execute the at least one instruction to:
receive a plurality of training images and a quality score set for each of the plurality of instances in each training image,
infer, from each training image, a quality score corresponding to each of the plurality of instances in the training image, and
obtain the third neural network by determining a weight of each of at least one layer included in the third neural network to reduce differences between the set quality scores and the inferred quality scores.
6. The computing device of claim 1 , wherein the image quality control parameters include an image quality control parameter for at least one of high-frequency component extraction, noise suppression sharpness enhancement, gray scale conversion, and hue correction.
7. A method of operating a computing device, the method comprising:
extracting features of an input image by using a first neural network,
recognizing a plurality of instances in the input image from the features of the input image by using a second neural network,
obtaining an image instance quality score map by predicting a quality score corresponding to each instance of the plurality of instances in the input image based on the features of the input image by using a third neural network, wherein the quality score corresponding to each instance varies depending on whether the instance is in motion or not in motion,
performing image quality processing differently for each instance of the plurality of instances in the input image by using the quality score corresponding to each instance of the plurality of instances in the input image by generating image quality control parameters for each instance in the input image by using the quality score corresponding to each instance in the input image, and performing the image quality processing independently for each instance in the input image by using the image quality control parameters corresponding to the instance, wherein as a value of the quality score corresponding to each instance decreases, an intensity of the image quality processing is applied more strongly, and as a value of the quality score corresponding to each instance increases, the intensity of the image quality processing is applied more lightly, and
generating an output image by merging the image quality-processed instances.
8. The method of claim 7 , wherein the image instance quality score map includes position information and the quality score of each instance of the plurality of instances in the input image.
9. The method of claim 7 , wherein the first neural network includes at least one convolutional layer.
10. The method of claim 7 , further comprising obtaining an instance map including a class corresponding to each instance, position information of each instance, and a mask indicating a pixel-unit region of each instance, by recognizing the plurality of instances in the input image from the features of the input image by using the second neural network.
11. The method of claim 7 , further comprising:
receiving a plurality of training images and a quality score set for each instance of the plurality of instances in each training image,
inferring, from each training image, a quality score corresponding to each instance of the plurality of instances in the training image, and
obtaining the third neural network by determining weights of at least one layer included in the third neural network to reduce differences between the set quality scores and the inferred quality scores.
12. The method of claim 7 , wherein the image quality control parameters include an image quality control parameter for at least one of high-frequency component extraction, noise suppression sharpness enhancement, gray scale conversion, and hue correction.
13. A computer-readable recording medium having recorded thereon a program that is executed to perform an operating method of a computing device, the operating method comprising:
extracting features of an input image by using a first neural network,
recognizing a plurality of instances from the features of the input image by using a second neural network,
obtaining an image instance quality score map by predicting a quality score corresponding to each instance of the plurality of instances in the input image based on the features of the input image by using a third neural network, wherein the quality score corresponding to each instance varies depending on whether the instance is in motion or not in motion,
performing image quality processing differently for each instance of the plurality of instances in the input image by using the quality score corresponding to each instance of the plurality of instances in the input image, by generating image quality control parameters for each instance in the input image by using the quality score corresponding to each instance in the input image, and performing the image quality processing independently for each instance in the input image by using the image quality control parameters corresponding to the instance, wherein as a value of the quality score corresponding to each instance decreases, an intensity of the image quality processing is applied more strongly, and as a value of the quality score corresponding to each instance increases, the intensity of the image quality processing is applied more lightly, and
generating an output image by merging the image quality-processed instances.