融合小波频域和马尔可夫链的数字传媒图像特征提取研究

Research on Feature Extraction of Digital Media Image by Combining Wavelet Frequency Domain and Markov Chain

针对现有数字传媒图像特征提取算法存在的边界控制不清晰、特征易丢失等问题,提出一种融合小波频域和马尔可夫链的数字传媒图像特征提取研究。利用小波频域算法去除数字传媒图像噪声信息(通过对不同子频图像的差异处理,最大限度地去除噪声信息,保护特征信息),以无噪数字传媒图像为基础,利用马尔可夫链构建边界闭合曲线,完成图像分割目的,基于Tiansi算子提取数字传媒图像每个分割区域的特征,从而实现数字传媒图像特征的快速、精准提取。实验数据显示:在不同实验工况背景下,提出方法应用后获得的数字传媒图像噪声比例最小值为1.02%,数字传媒图像分割综合评估系数最大值为0.96。

To address the issues of unclear boundary control and easy feature loss in existing digital media image feature extraction algorithms,a new feature extraction method that combines wavelet frequency domain with Markov chain was proposed.The wavelet frequency domain algorithm was used to remove the noise information of digital media images(through the difference processing of different sub-frequency images,the noise information is removed to the maximum extent and the feature information is protected).Based on the noise-free digital media images,the boundary closure curve was constructed by Markov chain to complete the image segmentation.Based on Tiansi operator,the features of each segmentation region of digital media image were extracted,so as to achieve rapid and accurate extraction of digital media image features.Experimental data show that under different experimental conditions,the minimum noise ratio of digital media image obtained by the proposed method is 1.02%,and the maximum comprehensive evaluation coefficient of digital media image segmentation is 0.96.