| Summary: | 碩士 === 國立中興大學 === 資訊科學與工程學系 === 96 === This thesis implements an image processing scheme for the extraction of regions of interest (ROI) in hand radiographs within a bone age assessment system. Features extracted from ROI images are used to describe the stage of skeletal development. More precise segmentations of ROI images can provide more accurate features for bone age assessment. The image processing of a bone age assessment system can be divided into two steps: extracting ROI images and analyzing ROI images. The first step of extracting ROI images is removing noises around the hand object. Then, a binary algorithm and a maximum component algorithm are applied to generate a binary image and eliminate small noises remained, respectively. Once the hand object is extracted, the axes of phalanges are depicted according to the tips of phalanges. Finally, ROI are located according to the information of intensities along the axes; and then the ROI images are extracted. In analyzing ROI images step, certain filters are used to smooth the texture and preserve the edges of bones first. Then an image segmentation procedure including two steps: two-means clustering and adaptive two-means clustering, is applied to segment regions of bone tissues and soft tissues. The two-means clustering generates an initial segmented image. Then the adaptive two-means clustering is applied to estimate final segmented image based on the intensity of original image and the initial segmented image.
In this thesis, a without background two-means clustering algorithm is proposed to provide a more precise initial classification, and a modified adaptive two-means clustering algorithm is proposed to complete the segmentation of ROI more accurately and more quickly. The experiment results reveal that the proposed scheme can greatly reduce the average error measures of segmentation. In addition, the reduction of the iteration numbers and processing time in the adaptive iteration segmentation step of the adaptive two-means clustering is another important contribution.
|
|---|
