Interactive graph cut method based on improved Gabor features for image segmentation

In this paper, an interactive color image segmentation approach is proposed. The method integrates color features with reduced Gabor features and then employs graph cuts method to obtain the segmentations. The reduced Gabor features are extracted by principal component analysis (PCA) from Gabor features. This way can overcome the problem that high-dimension Gabor feature vectors may contain some features irrelevant to the discrimination of texture. Besides, in order to fuse the color and texture features effectively, a new method is proposed to adjust the relative weight of these features. The experiments demonstrate the superior performance of the proposed method.