贝叶斯框架下图像细节模糊区域复原仿真

Restoration of Image Details in Fuzzy Regions under Bayesian Framework

针对当前图像模糊复原方法图像边缘信息丢失严重、复原系数低的问题,提出基于贝叶斯的图像细节模糊区域复原方法。对待复原图像进行预处理,在预处理后图像中选取前景种子点与背景种子点。在种子点中选取若干前景聚类中心与背景聚类中心,并定义分割图像至前景聚类中心及背景聚类中心之间的最短距离,得到待分割图像像素与前景、背景之间的相似度。将相似度值代入平滑性函数中,获得图像细节模糊区域分割结果,实现图像细节模糊区域的初步复原。计算由剩余噪声和图像原本信号组成的混合信号,得到贝叶斯准则。在噪声和图像信号均符合高斯分布的状况下,实现信号的线性调整。引入二参数拉普拉斯分布函数,完成贝叶斯估计,同时利用二参数得到图像细节模糊区域最终复原结果。实验表明,上述方法图像模糊复原系数最高为0. 98,图像边缘信息丢失率较低。所提方法具有很强的复原能力和保障图像完整性能力。

Due to serious image edge information loss and low restoration coefficient of current method,a method to restore image detail in fuzzy region based on Bayesian was proposed.This method preprocessed the image to be restored and selected the foreground seed point and the background seed point from the preprocessed image.Then,our method selected several foreground cluster centers and background cluster centers from seed points and defined the shortest distance from the segmentation image to the foreground cluster center and the shortest distance between background cluster centers,so as to obtain the similarity between the pixels of image to be segmented and the foreground and background.Moreover,the research introduced the similarity value into the smoothness function to obtain the result of the fuzzy region segmentation and realize'the initial restoration of image detail region.Furthermore,our research used the mixed signal consisting of the residual noise and the image original signal to obtain Bayesian criterion.Thus,the linear adjustment of signal was achieved under the condition that noise and image signal all conform to Gaussian distribution.Finally,we introduced the two-parameter Laplace distribution function to complete Bayesian estimation.At the same time,the two-parameter was used to obtain the final restoration result of image detail in fuzzy region.Simulations prove that the maximum coefficient of image restoration is 0.98 by using the proposed method and the information loss rate of image edge is low.The proposed method has a strong ability to recover image and guarantee the completeness of image.