CAJ | 학술논문

提出一种估计图像噪声的方法,该方法用混合高斯概率密度模型拟合图像的小波系数中最高频率子带的直方图,用EM算法估计模型的参数,选取其中最小的标准方差作为图像噪声标准方差.用该方法能准确地估计图像高斯噪声的标准方差,尤其当图像的噪声比较弱时,该方法比传统方法更准确.
제출일충고계도상조성적방법,해방법용혼합고사개솔밀도모형의합도상적소파계수중최고빈솔자대적직방도,용EM산법고계모형적삼수,선취기중최소적표준방차작위도상조성표준방차.용해방법능준학지고계도상고사조성적표준방차,우기당도상적조성비교약시,해방법비전통방법경준학.
A new method was proposed to estimate noisy image’s quality. This approach uses Gaussian mixture model to simulate the histogram of the highest subband’s coefficients in the wavelet domain. The parameters of this model can be calculated by EM algorithm. The smallest deviation of the Gaussian mixture model is selected as the noise deviation. The authors' proposed method can provide better result than the other traditional ones.