【摘要】 以颅脑CT图像为研究对象,提出了一种基于小波变换的自动标记非刚性配准所需对应特征点的算法。这种算法充分考虑了颅脑CT图像的像素点及其临域的纹理特征,通过进行小波变换建立对应于每个像素点的多分辨率小波特征向量,并以小波特征向量间的差异作为判别依据,在目标图像中标记非刚性配准所需的对应特征点。一系列的实验结果表明,这种基于小波变换的算法能够准确地在目标图像中标记出配准所需的对应特征点,可以作为基于特征的非刚性配准对应特征点自动标记的参量之一。更多还原

【Abstract】 A algorithm of automatic correspondence detection in brain CT image based on wavelet transform is presented.Adequately the algorithm thinks about the grain feature of pixel and its neighborhood in brain CT image,subsequently constructs a multi-resolution wavelet-based attribute vector(WAV) of the pixel.The WAV’s difference is used to detect the correspondence pixel in object image.Experimental results show the algorithm based on wavelet transform can detect correspondence accurately in object image,it would be one of parameters of automatic correspondence detection for deformable registration based on features.更多还原