Segmentation of Compressed High Speed Video Images of Vocal Folds

Abstract

High-speed endoscopic video images (HSV) of vocal folds contain important information for the analysis of vocal folds and their surroundings. However, HSV images have high data sizes and therefore require images to be compressed to reduce transmission times as well as high storage need. In this study, the effect of compression of HSV images on segmentation performance is investigated. First, HSV images are compressed in JPEG and JPEG-2000 formats. Subsequently active contour and region growing based segmentation of manually masked images are performed. As a result, it is shown that for similar segmentation performance, images can be reduced to approximately 1.77% of the data sizes in JPEG compression compared to the original images, and to 1.08% in JPEG-2000 compression.