基于Radon和伪Fourier-Mellin变换的图像退化不变性识别方法

Image Degradation Invariant Recognition Method Based on Radon and Pseudo-Fourier-Mellin Transforms

图像退化不变特征提取是图像识别中的关键技术,针对图像模糊、平移、拉伸和旋转退化等问题,本文提出了一种基于Radon和伪Fourier-Mellin变换的图像退化不变性识别方法。首先,介绍了相关的数学背景包括图像模糊的定义、Radon变换和伪Fourier-Mellin变换等;并讨论了基于Radon和伪Fourier-Mellin变换的图像组合退化不变特征,包括几何变换不变特征和模糊不变特征;最后,为验证所述方法的有效性,本文从图像不变性识别能力、计算复杂度两个方面进行了实验验证并与现有其他方法进行了比较。实验结果表明:本文所述的图像退化不变性识别方法在保证不增加计算复杂度的前提下,其识别能力和抗噪能力具有比较明显的优势。

Image degradation invariant feature extraction is the key technology of image recognition. In order to solve the degradation problems of image fuzzy,translation,stretching and rotating,a kind of image degradation invariant recognition method is put forward based on Radon and Pseudo-Fourier-Mellin transform in this paper. Firstly,the related mathematical background including the definition of image blur,Radon transform and pseudo Fourier-Mellin transform are introduced. And the image degradation invariant features including geometric transformation characteristics and fuzzy invariant characteristic are discussed based on Radon and pseudo Fourier-Mellin transform. Finally,the image recognition ability and computational complexity experiments are carried on to verify the effectiveness of the described method and it is compared with the existing methods such as GMI,CMI. The experimental results show that the image degradation invariant recognition method described in this article has obvious advantages on recognition and anti-noise ability without increasing computational complexity.