In order to improve the compensation effect for missing visual features of specific virtual images,a multimodal deep learning based automatic compensation method for visual features of specific virtual images is proposed.Firstly,by combining principal component analysis(PCA)dimensionality reduction method with structural tensor non local total varia-tion method,redundant information is removed and denoised from the image.Secondly,deep learning is carried out through sparse autoencoder and convolutional pooling techniques to obtain color,depth,grayscale,and 3D surface normal features of images under multimodal conditions.Finally,automatic compensation of specific virtual image visual features is achieved through convolutional neural networks.The test results show that this feature compensation method can improve image clar-ity through preprocessing,and the quality of image visual feature automatic compensation is high.
关键词
多模态/特定虚拟图像/视觉特征自动补偿/图像去噪/卷积神经网络
Key words
multimode/specific virtual images/automatic compensation of visual features/image denoising/convolution neural networks
维普
万方数据