基于多尺度多方向Gabor变换的Tsallis熵阈值分割方法

为了能在统一框架内处理无模态、单模态、双模态或者多模态直方图情形下的自动阈值选取问题，该文提出一种基于多尺度多方向Gabor变换的Tsallis熵阈值分割方法(MGTE)。该方法先通过Gabor变换得到多尺度乘积图像，然后利用内外轮廓图像从多尺度乘积图像中重构1维直方图，并在重构1维直方图上采用Tsallis熵计算模型来选取4个方向Tsallis熵取最大值时对应的阈值，最后对4个方向的阈值进行加权求和作为最终分割阈值。将提出的方法和5个分割方法在4幅合成图像和40幅真实世界图像上进行了实验。结果表明提出的方法虽然计算效率不占优势，但它的分割适应性和分割精度有明显的提高。

Tsallis Entropy Thresholding Based on Multi-scale and Multi-direction Gabor Transform

To deal with automatic threshold selection issue in non-modal, unimodal, bimodal or multimodal situations within a unified framework, a Tsallis Entropy thresholding segmentation method based on Multi-scale and multi-direction Gabor transform (MGTE) is proposed. The multi-scale product image is first obtained by the Gabor transform and then the inner and outer contour images are used to reconstruct the one-dimensional histogram from the multi-scale product image. Based on the reconstruction of the one-dimensional histogram, the Tsallis entropy calculation model is utilized to select 4 thresholds by maximizing Tsallis entropy in 4 different directions, and finally the weighted sum of the 4 thresholds is used as the final threshold. The proposed method is compared with 5 segmentation methods on 4 synthetic images and 40 real-world images. The results show that the proposed method has no advantage in computational efficiency, but its adaptability and segmentation accuracy are significantly improved.