| Abstrak/Abstract |
Raster-scan photoacoustic imaging often faces challenges with low resolution and extended acquisition times, limiting its effectiveness in biomedical applications. Traditional interpolation methods, such as nearest-neighbor, Bilinear, and Bicubic, do not fully address these issues, resulting in residual blurring and artifacts. This study investigates the use of Enhanced Super-Resolution Generative Adversarial Networks (ESRGAN) to enhance the resolution of raster-scan photoacoustic images. We compare ESRGAN with conventional
interpolation techniques to assess improvements in image quality. Our analysis, based on metrics including Full Width at Half Maximum (FWHM), Laplacian Variance, Edge Density, and Spectral Energy, shows that ESRGAN outperforms traditional methods. ESRGAN achieves a lower FWHM, indicating finer detail and sharpness, while conventional methods, which are nearest-neighbor, bilinear, and
bicubic, exhibit higher FWHM values and more blurring. Visual inspections confirm that ESRGAN images are significantly clearer and more detailed compared to those produced by traditional methods. These findings highlight ESRGAN's effectiveness in enhancing the image quality of raster-scan photoacoustic imaging |
