【摘要】 针对图像中存在的噪声,在同质性分割基础上提出了一种新的图像滤波方法。该方法首先利用形态学算子描述图像的不连续性,重新定义同质性,提高了图像的抗噪性。然后利用统计得到的同质直方图,并通过峰值寻找算法得到波峰、波谷,进而对图像进行分割。最后计算不同子区域的灰度均值,对满足条件的子区域进行合并得到滤波图像。实验结果表明,与其它滤波方法相比,该方法在滤除噪声的同时更好地保留了图像的细节和边缘信息,而且对含有不同噪声密度的图像均能得到较好的滤波效果。更多还原

【Abstract】 Aiming at the noises in image, a new image filtering approach based on homogeneity segmentation was proposed. This approach firstly described the discontinuity by using morphological operations and redefines homogeneity, improved the image’s anti-noise property. Then taking advantage of homogeneity histogram which was counted to gain the peaks and valleys by using peak-finding algorithm, the image into several small parts was divided. It lastly calculated the mean gray-scale value of different sub-region, combinations those areas which satisfied restrained conditions, and got the filtered image. The experimental results show that the proposed method effectively removes the noise of the image, and is better to reserve details of the image and the edge information compared with other methods. Moreover, the filtering effects are better than others for different noise density in the image.更多还原