在直接体绘制的过程中,传输函数的设计是一个关键的步骤,其对最终的绘制质量有着决定性的影响。超声图像有分辨率低、对比度低、充满乘性的斑点噪声等特点,因此很多适用于CT图像的传输函数设计方法并不适用于超声图像。本文通过对超声图像的特点进行研究,给出了一种基于局部纹理统计特征的传输函数设计方法。对超声图像序列在三维空间中提取出每个点的局部纹理统计特征,对这些特征进行聚类分析后,根据每个点所属的类和其到相应聚类中心的距离对其分配包括RGB值和透明度在内的光学属性。由于采用了统计特征作为分类特征,相较于使用梯度特征的传输函数设计方法,我们的方法能够更好地适应超声图像的特性,得到较好的绘制结果。
The design of the transfer function is a critical step in direct volume rendering process. It plays an important role for the final rendering quality. The ultrasound images have such characteristics as low resolution, low contrast, full of speckle noise, so some common transfer function design methods that work well when applied to CT images can not work when in ultrasound images case. Based on the study of the ultrasound image characteristics, this paper demonstrates a new transfer function design method based on local texture statistical features. We extract local texture statistical feature for each voxel in volume dataset and then classify all points. After that, optical properties including RGB values and transparency are allocated to each point based on the corresponding cluster center and the distance between the point to that center. The use of local texture statistical feature makes our method more suitable for ultrasound images than other methods.