三维动画视频关键帧图像信息提取方法研究

Study on image information extraction method of 3D animated video

在提取关键帧图像信息中,由于对特征的分析结果精度较低,导致信息提取结果与实际情况存在较大偏差,针对此,提出三维动画视频关键帧图像信息提取方法研究。首先分析了关键帧图像所在尺度空间的极值,利用三维动画视频关键帧图像的尺度空间函数和高斯函数之间的关系,计算相邻尺度上任意两个连续高斯图像的差值,计算得到差分高斯图像,根据采样点与图像区域相邻点关系,确定极值点;结合不同类型不稳定点的属性去除不稳定点后,根据特征点方向信息提取三维动画视频关键帧图像特征点描述符;最后在随机抽样一致性算法的辅助下,通过对图像平移、旋转和尺度变换矩阵求解,得到对应的图像信息。在测试结果中,设计方法对图像颜色信息中亮度信息的提取结果与实际数据拟合值始终在0.99以上,色度和饱和度的拟合值也始终稳定在0.89以上。

In the extraction of key frame image information,due to the low accuracy of the feature analysis results,there is a large deviation between the information extraction results and the actual situation.For this purpose,the study of the image information extraction method of key frames of 3 d animated video is proposed.First,the extreme values of the scale space of the keyframe images are analyzed,Using the relationship between scale spatial functions and Gaussian functions of 3 D animated video keyframes,Calculate the difference between any two consecutive Gaussian images on adjacent scales,Calculated to obtain the differential Gaussian images,According to the relationship between the sampling points and the adjacent points of the image area,Determine the extreme value point;After removing unstable points by combining the properties of different types of unstable points,Extracting the feature point descriptors of the 3 D animated video keyframe image according to the feature point direction information;Finally,with the assistance of the random sampling consistency algorithm,By solving the image translation,rotation,and scale transformation matrices,Get the corresponding image information.In the test results,the fitting results of extracting the brightness information in the image color information by the design method and the actual data are always above 0.99,and the fitting value of the color degree and saturation is always stable above 0.89.