基于多源融合FCN的图像分割

图像分割任务中，传统的基于人工设计特征方法工作量大、复杂度高、分割割精度较低，现有的基于全卷积神经网络(Fully Convolutional Networks，FCN)的方法在分割边缘上不够精细。为了提高图像分割算法的分割精度，提出基于多源融合的全卷积神级网络模型，输入图片经过Sobel算子提取边缘特征获得特征矩阵，与RGB和灰度图像一起作为输入，将传统全卷积网络拓展成具有多种输入源的分割模型。在PASCAL VOC 2012图像分割数据集上进行实验验证，结果显示该模型提高了图像分割的精度，具有良好的实时性和鲁棒性。

Image Segmentation Based on Multi - source Fusion FCN

In image segmentation, the traditional methods based on the manual feature extraction are labor intensive and complex, while the segmentation precision is low. The existing methods based on the fully convolution neural network are not fine enough on the edge of the segmentation. In order to improve segmentation accuracy, this paper proposes a full convolution neural network model based on multi-source fusion. The sobel operator is used to extract the feature of the image edge.And with the RGB and grayscale images as input, the traditional convolution network is extended to a multi-input segmentation model. Experiments on the PASCAL VOC 2012 image segmentation data set show that the model improves the accuracy of image segmentation, meanwhile, the model is real-time and has a good ability of robustness.