CAJ | 학술논문

研究了基于过完备字典下稀疏表示框架的图像去噪问题,基于 Stein 无偏估计提出一种自适应阈值选择算法.在一阶可导收缩函数的基础上,推导了阈值选择的优化目标函数;证明该函数是关于阈值的凸函数,用黄金分割法搜索其全局最小值.该算法选择的阈值接近峰值信噪比-阈值曲线的极大值点.通过对 Lena 和 Barbara 图进行去噪实验,验证了算法的优越性.
연구료기우과완비자전하희소표시광가적도상거조문제,기우 Stein 무편고계제출일충자괄응역치선택산법.재일계가도수축함수적기출상,추도료역치선택적우화목표함수;증명해함수시관우역치적철함수,용황금분할법수색기전국최소치.해산법선택적역치접근봉치신조비-역치곡선적겁대치점.통과대 Lena 화 Barbara 도진행거조실험,험증료산법적우월성.
@@@@Under sparse representation framework, image denoising problems are investigated over over-complete dictionary and a novel adaptive threshold selection method is presented based on Stein unbiased risk estimator. Based on the one order deriv-able shrinkage function, the optimal objective function about threshold selection is derived and it is shown to be convex function on threshold, and then its global minimum is searched by golden section method. The choice of threshold is closer to the peak signal to PSNR-threshold curve’s maxima. Lena and Barbara are used in experimental simulation, and the results demonstrate the superior performance of the algorithm.