奇异值分解与傅里叶变换相结合的模糊图像分类方法

Blurred Image Classification Method Combining Singular Value Decomposition and Fourier Transform

摘要: 提出了一种基于图像奇异值分解与傅里叶变换相结合的自动区分运动模糊与离焦模糊图像的方法。首先，剔除图像前几个奇异值加权的映射基进行傅里叶变换。其次，以频谱中心为中点取频谱图中合适大小的特征区域，用合适的阈值对该区域二值化。然后，对二值化后的图像进行闭操作与孔洞填充等形态学操作，提取特征连通域。最后，获取特征连通区域的长宽比，在划定阈值后，以长宽比的大小来区分图像的模糊类型。结果表明，该方法具有很好的适用性与模糊识别准确率。

Abstract: A method for automatically distinguishing motion blur and defocused blur images based on com-bining image singular value decomposition and Fourier transform is proposed. First, eliminate the first few mapping bases of image singular value weighting, and do Fourier transform. Secondly, the feature area of the appropriate size in the spectrogram is taken with the center of the spectrum as the midpoint, and the area is binarized with an appropriate threshold. Then, the binarized image is subjected to a morphological operation such as a closing operation and a hole filling to extract a feature connected domain. Finally, the aspect ratio of the feature connected region is obtained, and after the threshold is demarcated, the blur type of the image is distinguished by the aspect ratio. The results show that the method has good applicability and blurred recognition accuracy.