University of Electronic Science and Technology of China

Guilin University of Electronic Technology

Hangzhou Institute of Technology, Xidian University

Junzheng Jiang

Jiang Qian

Haitao Lyu

Minfeng Xing

Wanyuan Cai

Tingfang Tan

An image foreground-background segmentation method and system based on sparse decomposition and graph Laplacian regularization are disclosed. Firstly, an image is divided into a plurality of non-overlapping image blocks; Then, a foreground-background segmentation model of the image is established according to the image blocks; An image segmentation problem is divided into several sub-problems, which are solved by iteration; Finally, after the iteration, solutions of the problem are obtained; The obtained solutions are respectively matrixed and patched to obtain a foreground image, which is a foreground image of the whole image. The image foreground-background segmentation method uses the linear combination of graph Fourier basis functions to better represent the smooth background region. In addition, the graph Laplacian regularization is used to characterize the connectivity of foreground text and graphics while keeping sharp foreground text and graphics contours. The experimental results show that this method has better foreground-background segmentation effect.

Image foreground-background segmentation method and system based on sparse decomposition and graph Laplacian regularization

Provide is a novel mixed-noise removal method for HSI with large size. First, the underlying structure of the HSI is modeled by a two-layer architecture graph. The upper layer, called a skeleton graph, is a rough graph constructed by using the modified k-nearest-neighborhood algorithm and its nodes correspond to a series of superpixels formed by HSI segmentation, which can efficiently characterize the inter-correlations between superpixels, while preserving the boundary information and reducing the computational complexity. The lower layer, called detailed graph, consists of a series of local graphs which are constructed to model the similarities between pixels. Second, based on the two-layer graph architecture, the HSI restoration problem is formulated as a series of optimization problems each of which resides on a subgraph. Third, a novel distributed algorithm is tailored for the restoration problem, by using the information interaction between the nodes of skeleton graph and subgraphs.

Hyperspectral image distributed restoration method and system based on graph signal processing and superpixel segmentation

A noise suppression method and system for Inverse Synthetic Aperture Radar micro-cluster objects using a generative adversarial network (GAN) are provided. The method includes: constructing the GAN, including a generator and a discriminator; obtaining and inputting noisy simulation data into the generator to obtain a first output, comparing the first output with noiseless simulation data to obtain a first generator loss, inputting the first output and the distribution function into the discriminator for denoising discrimination to obtain a first discriminant result, and determining a second generator loss according to the first generator loss and the first discriminate result; and obtaining measured data and inputting the measured data into the generator to obtain a second output, inputting the second output to the discriminator to obtain a second discriminant result, and determining a generator loss according to the second generator and the second discriminate result.

Noise suppression method and system for inverse synthetic aperture radar micro-cluster objects using generative adversarial network

A moving target focusing method and system based on a generative adversarial network are provided. The method includes: generating, using a Range Doppler algorithm, a two-dimensional image including at least one defocused moving target, as a training sample; generating at least one ideal Gaussian point in a position of at least one center of the at least one defocused moving target in the two-dimensional image, to generate a training label; constructing the generative adversarial network, the generative adversarial network includes a generative network and a discrimination network; inputting the training sample and the training label into the generative adversarial network to perform repeated training until an output of the generative network reaches a preset condition, to thereby obtain a trained network model; and inputting a testing sample into the trained network model, to output a moving target focused image.

Growing community of inventors

Guilin, China

Junzheng Jiang