基于改进视觉注意模型的显著区域区检测

Region of Interest Detection Based on Improved Visual Attention Model

文中在专家学者提出的视觉注意模型的基础上,提出一种新的视觉注意模型改进算法。首先,采用改进后的Itti模型来计算图像视觉显著性,从底层特征,包括亮度、方向、颜色,然后建立小波金字塔,利用中央周围算子作差运算,得到颜色、亮度和方向每个特征的关注图,然后通过归一化算子,融合成Itti模型下的显著图;接着,借鉴了MA和Zhang等提出的显著图模型算法的一些方法,从全局的角度计算显著图;最后将以上得到的显著图进行归一化融合。

In this paper, propose a new visual attention model improvement algorithm based on the visual attention model proposed by ex- perts. First,the improved Itti model is used to compute the visual saliency of image, from low-level features including color, brightness and direction, then construct a wavelet Pyramid,using the central peripheral operator for difference operation to attain the concern graph of each feature of color,brightness and direction, and then through the normalized operator, merging into the saliency map of Itti model. Next,drawing on the experience of algorithm proposed by MA and Zhang ere, calculate the saliency map from a global perspective. Final- ly the saliency map above are normalized fusion.