中文摘要：

针对目前织物沾水等级评定方法不能有效消除噪声干扰和减弱光照不匀影响,提出L0梯度最小化和拉普拉斯特征映射相结合的方法解决上述问题。根据织物沾水图像的边缘信息,对图像中的像素值沿x、y方向求导,利用非零梯度个数约束整体平滑程度,并保持边缘不被平滑;然后把图像转化为CIELAB模式消除光照影响,计算邻域内颜色相似度;根据拉普拉斯特征映射理论计算其广义特征向量,组成颜色平缓过渡的图像,并使用模糊聚类算法对去噪后图像进行聚类分割。实验结果表明,该方法能够有效实现织物沾水图像的去噪处理,得到正确分割的织物沾水区域。

英文摘要：

Current detection method of fabric wetting performance cannot eliminate the noise interference and suppress the influence of uneven illumination. Thus,this paper proposes the combination of L.gradient minimization and laplacian eigenmaps for image smoothing in order to remove the above problems. According to the edge information of fabric wetting image, the derivative of pixel value in the image is figured out along the ＊ and y directions. The number of non-zero gradients is used to constrain the whole smoothness degree,and the edge is kept not to smoothen. Then, the image is transformed into CIELAB mode to eliminate the effect of illum in a tio n, and color similarity in the neighborhood is calculated. Meanwhile, the generalized egienvectors are calculated according to laplacian eigenmaps theory to consist of a piecewise smoothing image. F in ally, the clustering segmentation is carried out for the denoised image with fuzzy clustering. The results show that this method can effectively achieve denoising treatment of fabric wetting image,and gain the correct wetting region.