自适应权值调整GS图像分割算法

Weights Adjusting GS Segmentation Algorithm

近期国内外分割算法研究表明,当目标局部边缘性质相差较大时,局部自适应分割算法较全局分割算法可以取得更好效果。为了快速地进行GS(GreedySnake)图像分割,提出了一种不规则形状自适应图像分割算法,用于图像目标物体的边缘检测,同时基于伪逆算法,提出了一种自适应调整参数的方法,该方法保留了算法的反馈机制。在系统动态仿真中,为了避免动态边缘的停滞,新算法继承了greedySnake算法的能量公式,同时根据附近目标轮廓边缘及其周围测试点的性质,通过调整其权值参量来达到调整局部特性的目的,以便使轮廓自适应地逼近目标边缘。计算机仿真结果表明,将新算法模型用于捕捉多种目标物体的边缘,可较其他Snake算法取得较为良好的效果。

The recent researches have improved that local adaptive segmentation is particularly more attractive than the fully automatic segmentation when the property of the object＇s local boundary is not similar. For improve the segmentation speed, This article describes a novel approach to the self-adaptable segmentation of irregular objects in an image. The algorithm is based on Moore-Penrose operator. With adaptable energy function parameters, the Greedy Snake is attracted to boundaries by use of a direct feedback mechanism （ Greedy Snake）. To avoid undesirable local minima, every energy function＇s weight is adaptable according to the test point＇s property nearby, and a suitable local convergent algorithm is proposed which enables snakes to converge to target boundary points. Through computation simulation, the paper proves that the proposed approach is capable of inheriting the characters of the Greedy Snake algorithm, through adjusting the weight vector of the energy function, the new model changes the local character of the Snake, and make it approach to the aim object＇ s boundary automatically. When applying the new model and traditional method to extract contours from various images, the new greedy snake model performs better than related snakes.