基于特征算子的RGB-D图像手分割算法

Feature Operator-basedhand Segmentation Algorithm of RGB-D Image

将人手势识别过程中使用的信息设计为不同算子,利用手部/非手部区域的相邻像素点的深度或颜色特征的连续性的原理,从手心作为起始点进行搜索并检测特征算子,保证了结果像素形成一个联通区域。算法解决了手部颜色和深度数据不均匀的问题,没有手必须离传感器最近的限制,可以实现复杂场景下的手分割。

Designing different operator for the feature that human＇s gesture recognition, based on the principle that the color and depth of adjacent pixels of hand have continuity, we present an algorithm which search and detect the feature opera tor from the palm as the starting point center, it ensure the results is connectedarea of pixel. Algorithm solvesthe problems of uneven color or depth of hand and avoids the limitation that hand must benearest to the sensor and can achieve hand seg- mentation of complex scenes.