基于双重轮廓演化曲线的相似图像组分割模型

The model for segmenting group of similar images based on dual contour evolutional curve

ACGS(Active Contours With Group Similarity)模型在CV模型的基础上结合了矩阵的低秩性约束,能较好地分割目标特征缺失或错误的相似图像组,但对于灰度不均的相似图像组分割效果较差。而双重轮廓演化曲线的图像分割水平集模型在LBF模型的基础上引入了目标内外两条轮廓曲线,很好地克服了LBF模型对于初始轮廓的敏感性,对于灰度不均的单张图像分割效果较好。受此启发,本文提出了基于双重轮廓演化曲线的活动轮廓模型来分割相似图像组。该模型首先结合LBF模型来更好地分割灰度不均的图像;其次利用ACGS模型的低秩性质来保持图像间的相似程度,从一定程度上改善了LBF模型在能量函数最小化时易陷入局部极小值的情形;最后引入目标内外的两条轮廓曲线,通过两曲线在演化过程中分别对局部像素的直接作用而产生间接的相互联系,从而有效地克服LBF模型对于初始轮廓的敏感性问题,使得该模型改善了对于灰度不均的相似图像组的分割效果。实验结果表明,与CV、LBF、ACGS以及双重轮廓演化曲线模型的分割结果相比较,本文模型对于灰度不均的相似图像组的分割效果具有优越性。

he ACGS （Active Contours With Group Similarity） model which is based on the CV model and combines with the constraint of the matrix＇s low-rank property can play a good role in segmenting the groups of similar images in which the features of the object is missing or misleading, but it performs poorly on the groups of similar images with intensity inhomogeneity. However, the active contour model with dual contour evolutional curve introduces two curves that are inside and outside the object respectively and can overcome the problem that LBF model is sensitive to initial contour＇s location and size, thus the model peforms well in segmenting an image with intensity inbomogeneity. Inspired by the two models above, we propose an active contour model based on dual contour evolutional curve for segmenting the groups of similar images. Firstly the model combines the LBF model for a better segmentation of inhomogeneous images. Secondly it uses the low-rank property in ACGS model to keep images＇ similar degrees and improve the situation of LBF model＇s local minimum problem to some extent. Finally the proposed model introduces the two curves so that it can efficiently overcome LBF model＇s initial contour sensitivity problem by the interaction between the two curves during the evolution. Therefore, the proposed model has a better performance on the groups of similar images with intensity inhomogeneity. The experimental results show that compared with the CV model, LBF model, ACGS model and the active contour model based on dual contour evolutional curve, the proposed model can get a better result for segmenting the groups of similar images with intensity inhomogeneity.