Research On The Algorithm Of Salient Object Detection Based On The DenseNet

With the rapid development of information technology,the era of massive images and video has come.The massive amount of information has presented new challenges to computer vision processing technology.For complex environments,humans can quickly search for areas of interest,the study of salient object detection based on the human visual attention mechanism has been a hot topic in the field of computer vision processing.The core of the algorithm for salient object detection is to construct a visual attention model to simulate the visual attention mechanism better.Current salient object detection models either leverage prior information,or use features from the bottom up,or a combination of both.The key to the bottom-up approach is to effectively extract features that are useful for salient object detection and use them appropriately.In this paper,we propose a new method based on DenseNet to detect the end-to-end salient objects regions in the image.It can make full use of the feature information extracted from the Convolutional Neural Network.The network is easy to train and requires no extra parameters,which is simple and fast.By an extensive evaluation on widely used test data sets,the experimental results show that our proposed method has better results for images with complex backgrounds.In order to obtain the region of salient objects,we also propose a image segmentation algorithm based on Graph Cuts with saliency.It not only uses the saliency map of the image,but also combines the original information of the image,and use SLIC reconstruct the energy function for segmentation.Experiments show that our segmentation algorithm can effectively extract the salient objects.