A Segmentation Scheme Based on Rayleigh Distribution Model for Extracting Glottal Waveform from High-speed Laryngeal Images

In this paper, a segmentation scheme that combines thresholding, morphological operation and region growing is proposed for tracing the vocal fold opening region from the high-speed laryngeal image sequences. Specifically, the proposed scheme contains three steps: 1), histogram based global thresholding; 2), morphological operation on the segmented binary output from step 1; 3), region growing process to refine the segmentation. The glottal area waveform, which is the glottis area as a function of time, is obtained on a frame by frame basis from the segmented region of the vocal fold opening, or glottis. Representative clinical image recordings from subjects with normal and pathological voices are processed to validate our approach