Research on Digital Media Image Processing and Enhancement System for Computer Deep Learning Vision

Combining the object-oriented technology and tree view function of VB6, a graphical control method based on tree view is presented. The information about the algorithm is wrapped in these graphical controls by means of wrapping. The boundary characteristics of the original image are extracted by filtering it based on the deep belief net. The mathematical model of deep belief net and its probability distribution function are established. In this project, the method of combining neural network transformation and sparse representation is proposed to carry out multi-feature fusion of image boundaries, and sigma function is added to the output layer of deep confidence network model to reduce the complexity of iterative training process. By extracting the shape gradient feature of the image boundary, the edge discrete feature is removed and the digital image boundary enhancement is realized. Experiments show that the improved image has better boundary characteristics. Under the same sample conditions, the number of iterative training steps is less, and the image edge processing effect is better than the traditional scheme.