In this paper a novel decision based iterative filter for the detection and elimination of salt and pepper noise is proposed. Effective decisions based upon the noise density of the image are used to filter out noise while maintaining finer details in an image. A fixed size window is used at each step to maintain maximum correlation throughout the filtering process. Additional pre-edge and post-smoothing processing are also presented to further enhance the quality of image. Rigorous analysis over Kodak benchmark dataset containing 24 natural images indicates an exceptional performance boost for medium to extremely high noise density when compared with state of the art filtering techniques. The proposed filter is tested quantitatively and qualitatively using benchmark parameters including peak signal to noise ratio, image enhancement factor and visual representation. Even at noise density as high as 90% and 95%, the proposed filter outperforms the exiting filters providing better edge detail, less blurring and low streaking effects.

在本文中，提出了一种新颖的基于决策的迭代滤波器，用于检测和消除盐和胡椒噪声。基于图像噪声密度的有效决策可用于滤除噪声，同时保持图像中更精细的细节。在每个步骤都使用固定大小的窗口，以在整个过滤过程中保持最大的相关性。还提出了附加的前边缘和后平滑处理，以进一步提高图像质量。对包含24个自然图像的柯达基准数据集进行的严格分析表明，与最先进的滤波技术相比，对于中等到极高的噪声密度，其性能都有显着提高。使用基准参数（包括峰值信噪比，图像增强因子和视觉表示。即使在噪声密度高达90％和95％的情况下，所提出的滤波器也优于现有滤波器，从而提供了更好的边缘细节，更少的模糊和低条纹效果。