基于卷积神经网络的显著性检测

Saliency detection based on deep convolutional neural network

显著性检测在图像处理领域应用广泛,当前显著性检测主要有自底而上与自顶而下及一些相关或改进算法,它们各有优势和缺陷。提出了一种基于卷积神经网络的显著性检测算法,利用卷积神经网络在图像处理方面强大的功能提取图像特征,进行特征融合,最后得到显著性图,用于显著性检测。将本文方法与传统的显著性检测方法进行对比,发现本文方法效果显著。

Saliency detection is widely used in the field of image processing. At present,the saliency detection is mainly based on the bottom and the top,also includes some correlation or improved algorithm,and each has both advantages and shortcomings. A saliency detection algorithm based on convolution neural network is proposed in this paper. It takes use of the great power of convolution neural network in image processing to extract image features and do feature fusion,and finally forms a saliency map which is used in saliency detection. Comparing this method with the traditional detection methods,it shows that this algorithm is more effective.