基于像素分类的像素值阈值噪声检测算法

Pixel classification-based noise detector with pixel value threshold

针对图像噪声常见的三种脉冲噪声模型,在像素差异均等性原则下,以窗口分块方式将图像像素分为噪声像素和非噪声像素子集,进而以噪声图像上下像素值阈值推导噪声类型判别式,得到一种非线性像素分类的图像噪声检测算法.以经典图像(Lena、Baboon)分别叠加上不同密度噪声进行噪声图像检测算法仿真测试,由检测精确性和正确性来评估所提出的检测算法的检测性能.仿真结果表明,该检测算法具有较好的检测效果,缺点是容易将部分非噪声像素识别为噪声像素.

According to three kinds of impulse noises in image processing, a novel technique is presented to classify pixels by pixel difference uniformity with blocks into noise pixel sets and non -noise pixel sets. Moreover, a novel nonlinear noise detector is proposed to identify the noise types by a discfiminance with up - down pixel thresholds in noise image. The presented algorithm is tested by a classical image（Lena） and image（Baboon） with adding a different density noise. The simulation results are evaluated by two indexes, named detecting accuracy and detecting correctness. The results indicate that the proposed algorithm has a favorable detecting performance, but its shortage is easy to mistake a non -noise pixel for a noise pixel.