基于小波域隐马尔可夫模型多尺度图像分割

提出了一种基于小波域自适应上下文结构的多尺度图像分割算法（JACMS）．该算法为了减小计算复杂度，采用隐马尔可夫半树模型和参数加权训练算法，得到了可靠的初始分割．为了获得较好的区域一致性和边缘准确性，在进行尺度间融合时，采用自适应的上下文结构分别应用于图像纹理均质区域和图像纹理边缘，保证了图像大致轮廓的准确性和可靠性，提高了分割后图像纹理边缘的精确度．对合成图像与航摄像片的实验结果表明，该方法的分割错误概率低于传统的基于小波域隐马尔可夫树模型的图像分割方法，且对真实图像得到了理想的分割效果．

Multiscale Image Segmentation Using Wavelet-Domain Hidden Markov Model

An image segmentation algorithm based on wavelet-domain, referred to as joint adaptive context and multiscale segmentation （JACMS） was developed. Towards achieving lower computational complexity, we proposed a half hidden Markov tree （ HMT ） model and a weighting training algorithm. The technique provided a reli- able initial segmentation when applied to image segmentation. In order to ensure high accuracy of both texture classification and boundary localization during the interscale fusion, adaptive context structures were applied to homogeneous regions and texture boundaries respectively. Experimental results of the segmentation for both synthetic images and aerial photo indicate that the approach has lower miss classified probability than traditional wavelet-domain HMTseg method, and it can achieve satisfying segmentation of real images.