Self-adaptive region growing algorithm to segment images of spectral imaging for TCM assessment

The spectral imaging method for assessing the traditional Chinese medicine (TCM) can evaluate the quality of the medicines and identify their authenticity by using their spectral images. It can also make the assessing procedure fast and non-destructive. In order to eliminate the influence of the background noise on the assessing results of TCM spectral images, extraction of ROI (region of interest) of the TCM spectral images is needed. Since the difference inherent in different kinds of TCM spectral images and the available fixedthreshold segmentation method can not meet the demand of selfadaptive segment, a self-adaptive region growing algorithm to extract ROI of the images is proposed based on the characteristic of TCM. This algorithm can automatically choose the seeds and threshold of the region growing process according to the gray-level histograms of the TCM spectral images, the connectivity among the pixels is taken into the consideration during the growing period, and the region-fill technique is applied to modify the cavity in ROI after the growing. It proves that this algorithm can extract ROI of the images automatically and precisely, the deviation of segmentation is less than 8%, its antinoise ability is good, and the nonsensical growing result is not appear.