区域活动模型彩色图像分割研究

Genetic Optimization of Regional Activity Model for Color Image Segmentation Research

在图像分割优化问题的研究中,在图像提取识别感兴趣的目标中存在噪声和重叠,造成质量不高。传统的区域活动图像分割需要将能量泛函转换为封闭曲线,造成了算法复杂度高,图像分割边缘信息容易丢失,质量效果差。为解决上述问题,提出了一种新的遗传算法优化区域活动轮廓模型图像分割算法。根据区域的活动模型中融入了含有梯度方向指示性信息的边缘停止函数,使曲线在非期望边缘处加速演化,然后采用遗传算法交叉,变异操作优化图像区域活动模型参数。仿真结果表明,提出的改进的方法可以有效的对图像进行分割,大大的提高了图像分割的精确度和效率,改善图像分割的质量。

Research image segmentation problem. In view of the traditional regional activity model for image segmentation, due to the need of the energy functional is converted to the closed curve gradient descent flow, resulting in a complex algorithm, image segmentation edge information is easy to lose. The paper put forward a new image segmentation algorithm based on genetic algorithm. Based on regional activity model, the algorithm includes an edge stopping function which contains gradient direction information indicative, so that the curve evolve aceeleratedly in a desired edge, Then the operations of crossover and mutation was used to optimize the parameters of image region activity model. The simulation results show that the proposed method can be effective for image segmentation, greatly improve the segmentation accuracy, and speed up the efficiency of image segmentation.