摘要
为了精确地提取动态图像特征,为动画设计师提供更全面、更准确的视觉信息,文中提出基于SIFT-GMLBP的动态图像视觉信息提取方法。以关键点为像素中心,采用局部二值模式(LBP),通过比较其与邻域的灰度值获取LBP码,实现动态图像局部纹理特征捕捉;根据网格化LBP(MLBP)进一步将动态图像中的像素邻域划分为多个网格,使每个网格产生一个LBP值,降低特征向量的维数;结合Gabor滤波器,通过多尺度和多方向的纹理分析,提取动态图像在不同频率和方向上的局部结构信息,整合所有Gabor滤波器响应图像的GMLBP特征,形成包含原始动态图像在不同尺度和方向上的丰富纹理信息的特征向量。实验结果表明:该方法提取的关键点数量和分布非常合理,具有较高的稳定性和动态信息捕获能力,且该方法每秒能够处理高达30帧的图像。
A dynamic image visual information extraction method based on SIFT-GMLBP is proposed.It aims to extract dynamic image features accurately and provide animation designers with more comprehensive and accurate visual information.Key points are taken as pixel centers.By using local binary pattern(LBP)and comparing the grayscale values of the pixel centers and those of their neighboring regions,the LBP codes are obtained to capture local texture features of dynamic images.Furthermore,the pixel neighboring regions in the dynamic images are divided into multiple meshes by mesh-based LBP(MLBP),so that each mesh generates an LBP value and the dimensionality of the feature vectors are reduced.By combining Gabor filters and after multi-scaled and multi-directional texture analysis,the local structural information of dynamic images at different frequencies and directions is extracted.The GMLBP features of all of the Gabor filter response images are integrated to form feature vectors containing rich texture information of the original dynamic images at different scales and directions.The experimental results show that the number and distribution of the key points extracted with the proposed method are very reasonable.The proposed method is of high stability and dynamic information capture ability.In addition,it can process up to 30 frames of images per second.
作者
郑蔚
ZHENG Wei(Xinyang Normal University,Xinyang 464000,China)
出处
《现代电子技术》
北大核心
2024年第19期83-86,共4页
Modern Electronics Technique
