A COMPARATIVE STUDY OF DIFFERENT THRESHOLDING TECHNIQUES IN SEGMENTING POROUS GALLIUM NITRIDE IN FIELD EMISSION SCANNING ELECTRON MICROSCOPY IMAGE

Abstract

The quantitative analysis of porous gallium nitride (GaN) nanostructures is critical for understanding the optoelectronic properties and engineering the bandgap of photodetector devices. Yet, this task remains a challenge due to the complexity of pore morphology and the limitations of conventional image processing approaches in segmenting the porous region. In this study, field emission scanning electron microscopy (FESEM) images of photoelectrochemically etched GaN samples were examined to assess the effectiveness of different thresholding methods on segmenting the porous region by estimating the porous parameter. The thresholding under study are Otsu thresholding, manual thresholding, and adaptive thresholding. These methods were evaluated across multiple magnifications and etching durations to estimate their porosity and average pore diameter. The results indicate that manual thresholding achieved porosity errors as low as 17.87% in shorter etching samples, while Otsu’s thresholding yielded errors as low as 24.32% in longer porous samples. For average pore diameter estimation, manual thresholding similarly performed best at short etching durations with a 32.13% error, whereas Otsu’s thresholding was optimal for longer etching durations with only 2.76% errors. Regardless, the selection of the optimal thresholding method across different magnifications proved challenging, as no consistent pattern was observed. However, among the reliable estimates across magnification, most were found in 25kx and 50kx magnification images. Whereby at these two magnifications, many estimate produce a lower percentage error estimate of less than 20% compared to higher magnification images. This study further highlights the limitations of pixel intensity-based segmentation methods and the need for more advanced approaches for accurate quantification of porous GaN nanostructures.