一种基于视觉注意机制的感知物体提取算法

针对图像中感知物体提取问题，借鉴认知学对视觉信息表达的研究成果，提出一种感知物体的数学定义，并由视觉注意计算模型得到的视觉注意显著图选择区域增长算法的种子区域。建立图像中感知物体的马尔可夫随机场模型，通过度量图像中感知物体的显著性、边缘和同质性分布情况，最小化能量函数提取出图像中的感知物体。对含有飞机的图像中感知物体的提取实验，验证了提出算法的有效性、以及在认知上的合理性。

Extraction Algorithm of Perceptual Object Based on Visual Attention Mechanism

Extraction of perceptual object is a difficult and significant work in image analysis and understanding. It is convenient for further image processing, such as object recognition, scene classification. Aiming at the object extraction problem of perceptual object in image, a mathematic definition of perceptual object is proposed based on the visual representation research achievement. The seed of the region growing algorithm is selected by the salient map through the visual attention com putation model, the Markov random field model of the perceptual object is built. The object is extracted by minimizing the energy function by the measurement of the salient, edge, and homogeneous of the object in the image. The efficiency and rationality in cognition of the proposed algorithm are demonstrated by the experiment results for the images including planes.