Research Of Depth Estimation Algorithm Based On Machine Learning

As the development of science and technology,a great step has been made in the improvement of 3D display equipment,which makes realistic impressions for 3D images and videos.However,as the complex producing procedure of 3D filming techniques,which is also very expensive at the same time,3D resources are quite deficient in present.Applying depth estimation algorithm can not only greatly reduce the cost of 3D video production,but also can recycle existing huge number of 2D video resources,convert it into 3D video,supplement 3D resources,and effectively solve the problem of 3D resource shortage.Depth estimation algorithm based on machine learning has advantages in adaption,which makes no restriction in the types of scene.It can produce depth maps with high quality which are able to represent the depth changes of targets properly,has become a key research area of depth estimation.This paper mainly focus on depth estimation algorithm based on machine learning.The major work can be sumarized as follows:1.A depth estimation algorithm based on convolutional neural network is proposed.First,the framework of convolutional neural network called DepthNet is designed,which is used for depth estimation.Then by inputting the training data into DepthNet to update the network parameters,the mapping relationship between the original 2D image and the depth map is established.Next,the depth map of the target image is generated by using DepthNet to estimate the depth information of target image,and cross-bilateral filter is used to improve the effect of the depth map.Experimental results show that the depth estimation algorithm based on convolutional neural network can effectively improve the quality of generated depth maps.2.A visual dictionary training method for depth estimation is proposed.First,initial visual words are obtained after the training of machine learning in the depth image library.Then,the hard negative mining method is used to find the hard negative examples of visual word,which are used to train the classifier of visual words while updating visual words and depth information to establish the visual dictionary for depth estimation.Experimental results show that our training algorithm can mine visual words with obvious consistency in spatial structure from the depth image library.3.A visual dictionary-based depth estimation algorithm is proposed.A visual dictionary is constructed to detect visual words at multiple scales on the target image,so as to match the corresponding depth information,and finally depth estimation is finished.Experimental results show that the proposed algorithm can obtain depth maps with obvious scene structure,significant object boundary,accurate object position and continuous depth change.