【摘要】 纹理特征的提取是纹理分割的关键。在小波域内,对图像建立多尺度自回归滑动平均(MultiscaleAutoregressiveMovingAverage,简称MARMA)模型,并用小生境遗传算法对MARMA模型进行参数估计,将估计出的参数向量作为纹理分割的特征标准。不同纹理图像的MARMA模型的参数向量具有较好的可分性,这就保证了最终纹理分割的质量。更多还原

【Abstract】 Extracting texture feature is a key factor in texture segmentation.This paper proposes a novel method for texture segmentation using multiscale autoregressive moving average (MARMA) model in wavelet domain.Parameter vector is obtained by niche genetic algorithms,and then used as texture feature vector. Experiments indicate that parameter vectors of MARMA model for different textures are discriminable.Finally,practical examples illustrate the validity of the proposed method.更多还原