简体   繁体   English

在Matlab中从脸部提取嘴唇区域和参数,以用于嘴唇同步

[英]to extract lip region and parameter from a face in matlab so as to be used for lip synch

Hello everyone 大家好
i am doing my project in Matlab. 我正在Matlab中做我的项目。 For this i would be using a lip extraction algorithm which would extract the lip region from the given facial image. 为此,我将使用嘴唇提取算法,该算法将从给定的面部图像中提取嘴唇区域。 For that i have used color based segmentation by which i can obtain the lip region. 为此,我使用了基于颜色的分割,通过该分割,我可以获得嘴唇区域。 This gives me the Lip region. 这给了我唇部区域。 However i am not getting the lip parameters. 但是我没有得到嘴唇的参数。 Hence i would have to use a standard lip for lip sync. 因此,我将不得不使用标准的嘴唇进行嘴唇同步。 Ie irrespective of the type of lip a sketch might have,we would have to use the same lips. 即,不管草图可能具有哪种类型的嘴唇,我们都必须使用相同的嘴唇。 This technique wouldnt give me an appropriate and accurate result. 这种技术不会给我合适和准确的结果。
What i want to ask is, "Is there any algorithm or method by which i can use the same lip of the sketch and appropriately modify the same lip for each and every viseme(smallest visual element) depending on the sound phoneme" 我想问的是,“是否有任何算法或方法可以根据声音音素使用草图的相同嘴唇并为每个视位(最小的可视元素)适当修改相同的嘴唇”
Thanking you all in advance 提前谢谢大家

I wrote my master thesis on this subject. 我写了关于这个主题的硕士论文。 I used Active Shape Models to determine the exact 'shape' or position of the lips in order to reduce parameters. 我使用活动形状模型来确定嘴唇的确切“形状”或位置,以减少参数。

http://www.vision.ee.ethz.ch/teaching/sada/sadalink/reports/biwi_00349.pdf http://www.vision.ee.ethz.ch/teaching/sada/sadalink/reports/biwi_00349.pdf

It works fairly well, but still, robustness is a huge problem. 它工作得很好,但鲁棒性仍然是一个巨大的问题。 I wouldn't rely on color segmentation to extract the lip contours.. it seemed to work only for data sets with high resolution and ideal lighting conditions. 我不会依靠颜色分割来提取嘴唇轮廓..它似乎仅适用于具有高分辨率和理想照明条件的数据集。 Additionally, some people have very thin or skin-toned lips. 此外,有些人的嘴唇很薄或肤色。

Anyway, have a look at the references in the report, they should be helpful. 无论如何,请查看报告中的参考文献,它们将对您有所帮助。

声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM