Design of a noise-dependent shrinkage function in wavelet shrinkage of X-ray CT image
Naruomi Yasuda1, Yoshie Kodera
1School of Health Sciences, Gifu University of Medical Science, Seki-shi, Gifu, Japan. yasuda@nagoya-u.jp
Purpose:
Many shrinkage functions have been introduced and applied for the wavelet shrinkage denoising of computed tomography (CT) images. However, these functions have problems in continuity of functions and cause "shrinkage artifacts". Therefore, we designed a new and smooth shrinkage function using noise distribution.
Methods:
The proposed shrinkage function was designed under the following four conditions: (1) use of noise distribution, (2) shrunk coefficients having all ranges of amplitude, (3) function continuity, and (4) property of a function that is controllable by two parameters. The designed function was applied to phantom and chest CT images and denoising performance was compared with other functions.
Results:
In the proposed method, edge and pixel values were maintained when compared with previous functions, the occurrence of shrinkage artifacts was smaller, and high- quality denoised images were obtained.
Conclusions:
The proposed shrinkage function is effective for low-dose noisy CT images when using accurately selected parameters.
