Method and apparatus for compressing or restoring image

1. A method for restoring a compressed image by an electronic device, the method comprising:

receiving monochrome image data and low resolution color image data each generated from an original color image;

decoding the monochrome image data and generating a low resolution monochrome image;

decoding the low resolution color image data generating a low resolution color image;

processing the low resolution monochrome image and generating a high resolution monochrome image in accordance with a super resolution imaging neural network; and

generating a high resolution color image based on the low resolution color image and the high resolution monochrome image in accordance with a colorization imaging neural network,

wherein a part of an intermediate layer of the colorization imaging neural network is based on the low resolution color image or an intermediate resolution color image having a resolution improved from that of the low resolution color image, and

wherein the generating the high resolution color image comprises inserting the low resolution color image or the intermediate resolution color image into one channel of the intermediate layer or performing element-wise addition on a value of at least one channel of the intermediate layer and a value of a feature of the low resolution color image or the intermediate resolution color image.

2. The method according to claim 1 , wherein the monochrome image data and the low resolution color image data are video data which are synchronized with each other, and wherein the part of the intermediate layer of the colorization imaging neural network is based on the low resolution color image corresponding to the low resolution monochrome image of the monochrome image data.

3. The method according to claim 2 , wherein the part of the intermediate layer of the colorization imaging neural network is based on the low resolution color image corresponding to an intra frame of a group of pictures (GOP) including the low resolution monochrome image.

4. The method according to claim 2 , wherein the part of the intermediate layer is based on the low resolution color image corresponding to a previous frame of the low resolution monochrome image or the low resolution color image corresponding to a frame which is referenced by the low resolution monochrome image.

5. The method according to claim 1 , further comprising:

generating at least one chrominance image at an output layer of the colorization imaging neural network; and

generating the high resolution color image by synthesizing the chrominance image and the high resolution monochrome image.

6. The method according to claim 5 , further comprising:

up-sampling the chrominance image; and

generating the high resolution color image by synthesizing chrominance information of the up-sampled chrominance image and brightness information of the high resolution monochrome image.

7. A non-transitory computer readable recording medium which stores a computer program configured to allow a computer to execute the method according to claim 1 when the computer program is executed by the computer.

8. An apparatus for restoring compressed image, comprising:

a processor; and

a memory which is electrically coupled with the processor and is configured to store at least one of at least one code executed in the processor, a parameter of a super resolution imaging neural network, or a parameter of a colorization imaging neural network,

wherein when the memory is executed by the processor, the memory stores codes which cause the processor to decode monochrome image data and low resolution color image data generated from an original color image, generate a low resolution monochrome image and a low resolution color image, generate a high resolution monochrome image by processing the low resolution monochrome image in accordance with the super resolution imaging neural network, and generate a high resolution color image by using the low resolution color image and processing the high resolution monochrome image in accordance with the colorization imaging neural network, and

wherein the super resolution imaging neural network is coupled to and simultaneously trained with a lower resolution imaging neural network so as to generate a monochrome image having a first resolution using, as an input, a monochrome image having a second resolution generated from the monochrome image having the first resolution by the lower resolution imaging neural network of an image compressing apparatus, and the first resolution is higher than the second resolution.