基于水平集复杂背景下毒蘑菇图像的提取

摘要

Extraction of poisonous mushrooms image is a key step from image processing to image analysis and pattern recognition. As a technique to separate a foreground portion from the background, digital matting is prone to bring unassociated pixels of small areas.For images of poisonous mushrooms taken from outdoors often exist in complex backgrounds, three methods of image segmentation like Kmeans, digital matting and digital matting combining with level set are used. Experimental results show that by introducing level set into target image extraction, with automatic convergence, target image can be well extracted from complex background image. This can provide an effective foundation for identifying poisonous mushrooms.%毒蘑菇图像的提取是对其进行分析与模式识别的关键步骤。数字抠图作为一种把图像的前景部分从背景中分离出来的技术，存在繁琐和容易产生小区域不相关像素的缺点。针对野外采集的毒蘑菇图像背景复杂的特点，在MatlabR2010a环境下，采用Kmeans算法自动分割、数字抠图及数字抠图和水平集结合3种方法进行毒蘑菇图像分割。实验结果表明：通过引入水平集进行目标图像提取，结合水平集的自动收敛，在复杂背景下，能很好地提取目标图像，为进一步识别毒蘑菇提供有效的基础。