| Summary: | 碩士 === 臺中技術學院 === 資訊科技與應用研究所 === 98 === In this paper, the Contrast-Detail phantom (CD phantom) image was corrected and the information of holes & background were obtained automatically. Then this paper judged if they are recognizable objects according to the information of holes in square. This paper also proposed an appropriate automatic CD phantom image quality optimization system as a criterion of CD phantom image quality.
First in this paper, the square Plexiglas region and the test region were segmented form the CD phantom image. Then the four corners of the test region were located linearly, and the geometry distorted bilinear equation model and the bilinear pixel interpolation of rectangular test region were used to recover the image of test region.
In the test region, this paper would segment the squares and locate holes region. With the x-ray phenomena create parallax effect, the holes were be shifting by region shifting (RS), this method would locate holes region accurately. Finally, this paper calculated and gathered statistics of hole''s contrast, signal to noise ratio (SNR) and probability density relationship (PDR). To determine whether the holes can be identified and located with human eyes to make a criterion for CD phantom image quality quantitative analysis.
In this paper, the experiment used seven CD phantom images (Source: Taichung Veterans General Hospital Radiology); six images were used as training images and one image was used as a test image. Training images had been viewed and judged by professional physicians and threshold of holes’ contrast, SNR and PDR for each row could be obtained by training images. In order to verify the accuracy and availability of this method, the result derived by holes'' contrasts, SNR, and PDR were analyzed and compared with the results analyzed by the physician respectively. The experimental results showed that the method in this paper can obtain accurate location and information of holes, and also provide a simple and fast way to achieve that. It can be an effective criterion for determining CD phantom image quality.
|
|---|
