| Title |
Histopathology Image Super Resolution using Generative Adversarial Network |
| Authors |
박준현(Joon Hyeon Park) ; 선우명훈(Myung Hoon Sunwoo) |
| DOI |
https://doi.org/10.5573/ieie.2022.59.8.55 |
| Keywords |
Deep learning; CNN; Super-resolution; Histopathology; Medical image |
| Abstract |
With the development of image digitization technology, biopsies performed under a conventional microscope are captured and stored as digital images through a slide scanner. However, in order to take histopathology image taken with a slide scanner, a long recording time is required, and the stored images amount to several gigabytes per sheet, and additional storage space is required to manage them. In order to solve this problem, studies are emerging that apply the super-resolution technique of saving at low resolution and then restoring it to histopathology images. In this paper, a super-resolution histopathology image close to the original and visually natural is generated by adding the residual image generated using the generative adversarial network to the bicubic interpolated image. As a result of the experiment, a PSNR of 29.24dB was achieved when magnified by 4 times, and it was confirmed that an image close to the original can be generated qualitatively. |