Image processing apparatus using neural network and method performed by image processing apparatus

A method for processing an image, the method comprising:

• converting a three-dimensional input image into a two-dimensional input image, based on directions of a receiver and a scanline, which are used to obtain the three-dimensional input image;
• inputting the converted two-dimensional input image into a neural network learned in advance to change a parameter associated with the converted two-dimensional input image;
• and
• generating a three-dimensional output image in which a quality of the three-dimensional input image is enhanced, based on the changed parameter, wherein
• the generating includes:
• determining information lost from the two-dimensional input image, based on the changed parameter.

A method for processing an image, the method comprising:

• converting a three-dimensional input image into a two-dimensional input image, based on directions of a receiver and a scanline, which are used to obtain the three-dimensional input image;
• inputting the converted two-dimensional input image into a neural network learned in advance to change a parameter associated with the converted two-dimensional input image;
• and generating a three-dimensional output image in which a quality of the three-dimensional input image is enhanced, based on the changed parameter, wherein
• the generating includes:
• generating the three-dimensional output image including information lost from the two-dimensional input image, based on the changed parameter.

An image processing apparatus comprising:

• a memory configured to store a three-dimensional input image;
• and
• a processor configured to generate a three-dimensional output image in which a quality of the three-dimensional input image is enhanced, by using a neural network learned in advance,
• wherein the processor is configured to:
• convert the three-dimensional input image into a two-dimensional input image in receiver-scanline coordinates;
• input the converted two-dimensional input image into the neural network;
• determine a coefficient associated with the two-dimensional input image so as to adjust noise included in the two-dimensional input image, based on the output of the neural network;
• and
• generate the three-dimensional output image based on an output of the neural network.