非邻域均值滤波算法能够在有效地去除图像中的高斯噪声的同时保留图像细节信息,但是该算法对于压缩噪声的滤除效果不佳.而且在非邻域均值滤波算法中,需要预先设定滤波参数,该参数对滤波性能的影响很大.但是在实际的应用中,很难凭经验预先设定滤波参数.鉴于此,设计出一种无参块效应评估准则T,T可在无原图参考的情况下,估计压缩图像的图像块块效应强度.并应用此准则设计了一种新的基于非本地均值滤波的去块效应算法,该算法依据上述准则自适应的为每个图像块选取最佳滤波参数,从而实现逐块去块效应滤波.%Non-local means filter when applied on an input image, efficiently removes the Gaussian noise without over-smoothing the details. However, the non-local means filter is not suitable for removing the blocking artifacts. In addition, a filter parameter which significantly affects the filtering result needs to be selected properly. But this is not practical for most applications. In this paper, a no-reference metric T is proposed to measure the deblocking efficiency. And a novel patch-adaptive deblocking algorithm is proposed for removing image blocking artifacts based on non-local means filter using metric T. In the proposed algorithm, non-local means filter is applied block by block and T is used to control the degree of filtering adaptively for each block.
展开▼