基于超像素的物体似然概率计算模型研究

提出基于超像素建立物体似然概率模型来检测图像的显著区域。首先根据显著性原理和物体的自然属性分析影响物体显著度大小的因素；然后使用SLIC算法把图像分成K个超像素，并根据纹理、颜色、梯度特征信息建立不同准则下显著物体概率的计算模型:包括类内紧凑性、颜色空间分布估计以及边缘连续性；再结合细胞调节和指数函数的特征对每个准则下的显著物体概率组合得到物体的似然概率；最后利用该算法在较复杂的场景中对显著区域进行提取实验，证明该算法比其他算法更高效。

Calculating probability of objectness likelihood model based on superpixels

Establishing calculation model on the probability of saliency objectness likelihood based on superpixels was introduced to dectect image saliency. At first, factors which affected the size of saliency were analyzed according to the principle of saliency and natural characteristics; And then the SLIC algorithm was used to divide image into K superpixels; Next, according to the texture, color and gradient feature information, calculation models were established on probability saliency object under different rules: including compactness in class, color spatial distribution estimation and edge continuity; Moreover,integrating the probability of saliency object under each rule to get the probability of objectness likelihood according to the characteristics of combining with activity in cells responding to stimuli and exponential function; Finally some experiments extracting regions of interest from complex scenes prove that the proposed algorithm is more efficient than other algorithms.