欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
从最大后验概率估计和马尔科夫随机场出发，将图像梯度场的分布建模为混合加权的容许密度类，利用鲁棒统计学中的Hubber定理，导出了一类鲁棒性密度。建立了一类由L^2范数或L^1范数数据保真约束和鲁棒意义下的图像正则化项组成的噪声抑制变分模型。提出了谈类模型的基于梯度最速下降的有限差分算法。在Matlab集成环境下进行了六组不同噪声抑制变分模型的仿真实验，通过计算峰值信噪比和结构化相似指标给出了性能评价结果。
英文摘要：
Assuming that the gradient of images is a member of a class of hybrid weighted probability distribution, a class of least favorable distribution was obtained based on the Hubber theorem, Then a class variational functional for image de-noising was set up in the,form of robustness regularized term under constraint of data fidelity with L^1 norm or L^2 norm. A gradient descent flow for image denoising with a iterative finite difference algorithm was proposed. Simulated experiments were implemented for six de-noising models with respect to Gaussian noise and Laplace noise, and the performance evaluation was given through computing the PSNR （Peak Signal Noise Ratio） and SSIM （Structural Similarity Index Method）.
同期刊论文项目
同项目期刊论文
期刊信息
位置：