基于直线截距直方图的Arimoto熵或Arimoto灰度熵的食品图像分割

Line Intercept Histogram-based Arimoto Entropy or Arimoto Gray Entropy for Food Image Segmentation

摘要:

食品生产中涉及到的食品种类繁多且必须满足国家相关食品安全标准，为此要求食品图像分割方法必须速度快、准确性高、普适性强。利用基于二维Arimoto熵或二维Arimoto灰度熵的阈值选取方法对食品图像进行分割，算法复杂度高，难以满足实时性要求。针对这一问题，提出基于直线截距直方图的Arimoto熵或Arimoto灰度熵的食品图像分割方法。首先给出直线截距直方图的定义，然后根据此定义建立图像的直线截距直方图，最后计算基于此直线截距直方图的不同灰度级的Arimoto熵或Arimoto灰度熵，当该熵达到最大时，对应的灰度级即为图像的最佳分割阈值。针对此方法，对多种食品图像进行了大量的试验，通过与现有的基于一维和二维Arimoto熵、Arimoto灰度熵的分割方法对比，发现本文方法在综合提升算法速度和改善分割效果上，性能更优。

Abstract:

In food production, the numerous kinds of foods produced have to meet the applicable national food safety standards. Therefore, methods for food image segmentation should be rapid with high accuracy and high universality. In food image segmentation methods based on the threshold of two-dimensional Arimoto entropy or two-dimensional Arimoto gray entropy, the algorithm is highly complex. Hence, it is difficult to meet real-time requirements. To solve this problem, a method using line intercept histogram-based Arimoto entropy or Arimoto gray entropy for food image segmentation was proposed. First, the line intercept histogram was defined, which was followed by building of the line intercept histogram of images according to this definition. Finally, Arimoto entropies or Arimoto gray entropies of different gray levels in this line intercept histogram were calculated. When the maximum entropy was reached, the corresponding grayscale was determined to be the optimal image segmentation threshold. Several experiments were performed on different kinds of food images by using this method. Compared with the existing segmentation methods based on one-dimensional and two-dimensional Arimoto entropy and Arimoto gray entropy, the method proposed here can achieve better performance by increasing algorithm speed and improving segmentation results.