Abstract: For blurring fluorescence microscopic image with Poisson noise, Richardson-Lucy（RL） algorithm can obtain good restoration results. Serious ringing and noise amplification phenomena, however, is generally accompanied. Therefore a new image restoration algorithm is proposed based on traditional RL algorithm in this paper, which is mainly considered from two aspects. One is that a gain map constraint is imposed on RL method, thus the contrast of the image edge is reduced, which can suppress the ringing effect. The other is that the Gabor wavelet decomposition technology is used in every iteration, which can overcome the problem of noise amplification during the course of restoration. Furthermore, more direction salient features of the image can be extracted, which can effectively increase the clarity of the human vision. By appropriately choosing the reasonable parameters, the best restoration result can be obtained. The experimental results show the effectiveness and feasibility of the proposed algorithm.

Key words: image restoration, Richardson-Lucy algorithm, gain map, Gabor wavelet

摘要： 对于含有泊松噪声的模糊荧光显微图像，Richardson-Lucy（RL）算法具有很好的复原效果，但经常伴有严重的振铃和噪声放大等现象。针对这一情况，在传统RL算法基础上，提出了一种新的图像复原算法，主要考虑两个方面，一是通过施加增益图约束，进而降低图像边界的对比度来抑制阶梯效应。二是将Gabor小波分解技术引入到每次迭代中，克服了复原过程中的噪声放大问题，同时能够多方向地提取图像特征，增加视觉上的清晰度。通过对参数的合理选择，获得最佳的复原效果，实验结果验证了该算法的有效性和可行性。

关键词: 图像复原, Richardson-Lucy算法, 增益图, Gabor小波