基于 SVM 的模糊图像识别

目的研究如何准确识别清晰图像与不同程度的模糊失真图像。方法首先对图像进行特征提取,主要从离散余弦变换域内的频率系数统计特征、峰度值、颜色饱和度三方面进行。然后在不同程度的模糊图像库中,利用支持向量机分辨出模糊图像。结果基于上述3种图像特征的组合,非常适合用于描述图像模糊现象,并且运用支持向量机分类器可以较为准确快速地区分出高斯模糊图像和清晰图像。结论提取模糊图像具有表征性的特征,可应用于不同程度模糊图像的识别,且运用支持向量机分类结果准确度也较高。此方法可应用于图像处理前期,剔除有碍信息表达的模糊图像。

SVM-based Recognition of Blurred Image

To study how to accurately recognize clear images and blurred images at different degrees, three characteristics of images were firstly extracted from the following three aspects: statistical characteristic of frequency coefficient, and kurtosis coefficient and saturation in Discrete Cosine Transform. Then a blurred image could be recognized by support vector machine (SVM) from the library of blurred images at different degrees. The combination of the above three image characteristics were very suitable for describing images blur. And the Support Vector Machine could distinguish Gaussian Blur images and clear images very accurately and rapidly. Extraction of the typical characteristics of the blurred images can be applied to the recognition of images with different degrees of blur. And the accuracy of support vector machine classification is relatively high. Therefore, this method can be applied to the early stage of the images processing, to eliminate some blurred images that hinder information expression.