A hybrid-algorithm-based region of interest segmentation in THz imaging

摘要

Region of interest segmentation is essential for computer aided application of THz imaging. However, THz images isseverely degraded by motion blur, poor resolution and noise. A robust, accurate and time-saving algorithm is in dire needfor the ROI segmentation of THz images. Recently, ROI segmentation of THz-TDS images and passive THz images hasbeen widely studied. While the ROI segmentation of THz continuous wave (CW) image is still in its infancy. In thispaper, we proposed a hybrid ROI segmentation method for THz CW images. The hybrid method combines block match3D denoising, fuzzy c-means clustering, morphology operation and canny edge detection. The hybrid method isimplemented to two images acquired with a THz CW reflection imaging system. To evaluate the performance of ouralgorithm, we calculated the accuracy, sensitivity and specificity. As the result indicates, this h ybrid ROI segmentationmethod performs well for THz images.