Jiang Shengqin

Bio: Jiang Shengqin is an academic researcher. The author has contributed to research in topics: Sparse approximation & Brightness. The author has an hindex of 1, co-authored 2 publications receiving 12 citations.

Single image super-resolution reconstruction method based on depth component learning network

Abstract: The invention discloses a single image super-resolution reconstruction method based on a depth component learning network, which comprises the following steps: expanding a training sample image and carrying out region extraction and degradation operations to obtain corresponding high-resolution and low-resolution image training sets; a deep network with component learning structure being constructed. The network decomposes the input low-resolution image into global components, and then predicts the corresponding image in high-resolution space using the residual components extracted from the global components. Batch random gradient descent method and back propagation algorithm are used to train the constructed deep component network iteratively on the training set, and the model with optimized weights is obtained. Reconstruction of Low Resolution Images Using the Trained Component Networks; the reconstruction result is restored to the original color space, and the final output of the super-resolution reconstruction is obtained. The method of the invention not only can improve the quality of the reconstructed super-resolution image, but also can improve the operation speed of the model.

Single image super-resolution reconstruction method based on two-way alignment sparse representation

Abstract: The invention discloses a single image super-resolution reconstruction method based on two-way alignment sparse representation, comprising the following steps: (1) reading a color low-resolution image, and transforming the image from an RGB color space to an YCbCr color space; (2) primarily converting the image into an image of a target size through bicubic interpolation, performing a super-resolution reconstruction modeling operation based on two-way alignment sparse representation on the brightness component of the image after conversion, and iteratively solving the reconstruction model through an iterative shrinkage thresholding algorithm to get the optimal estimated value of the brightness component of the high-resolution image; and (3) transforming the image from the YCbCr color space to the RGB color space, and getting the final output of super-resolution reconstruction The method not only can improve the quality of reconstructed images, but also is superior to the traditional methods in robustness

Image reconstruction model training method and image super-resolution reconstruction method and device

Abstract: The invention discloses an image reconstruction model training method and an image super-resolution reconstruction method and device, and belongs to the technical field of image super-resolution. Themethod comprises the steps of obtaining a sample set by preprocessing an image, establishing an image reconstruction model for image super-resolution reconstruction; using the sample set to train andtest the image reconstruction model; in the image reconstruction model, using a feature extraction network for performing feature extraction on a low-resolution image and inputting into a first residual network, wherein the m cascaded residual networks are respectively used for carrying out feature extraction on an output image of a previous network and then superposing the output image with the image, m attention networks are respectively used for extracting images of a region of interest from the output images of the m residual network, and an amplification network is used for fusing and amplifying the output images of the attention networks and the m residual networks, so that the output images and the image subjected to bicubic interpolation amplification are fused by the first fusionlayer. According to the present invention, the visual effect of the reconstructed image can be effectively improved.

Super-resolution image reconstruction method based on lightweight network

Abstract: The invention discloses a super-resolution image reconstruction method based on a lightweight network. The method mainly comprises a network lightweight design module and a network lightweight designmodule, and the network lightweight design module is mainly used for improving an original EDSR network structure by utilizing a ShuffleNet unit structure designed by the invention, so that the structure is simplified, the network parameters are greatly reduced, and the storage pressure is reduced; the network quantization module is mainly composed of a network pruning part, a weight sharing partand a Huffman coding part, through combination of the three parts, network parameters are quantized, and the coding mode is changed, so that the network parameter quantity is greatly compressed, and the calculation speed is increased. The network structure is improved on the basis of an existing image super-resolution reconstruction network, optimization is carried out by combining various deep compression methods, the effect after image reconstruction can be effectively guaranteed, and meanwhile the effects of few network parameters, high processing speed and high portability are achieved.

Image super-resolution reconstruction method based on high and low frequency information decomposition

Abstract: The invention discloses an image super-resolution reconstruction method based on high and low frequency information decomposition, and the method comprises the following steps of carrying out the datapreprocessing on a super-resolution data set, and obtaining an LR-HR image block pair needed by the training of a neural network; carrying out high and low frequency information decomposition on theHR image block G; constructing a convolutional neural network model; training the constructed convolutional neural network model by using the training sample data generated in the previous step to obtain an optimized convolutional neural network model; and inputting an LR image by using the optimized convolutional neural network model to generate a corresponding HR image. According to the method,an optimized neural network structure is designed, the low-frequency and high-frequency information of the HR image is effectively generated, and a high-resolution image is better reconstructed.

An image processing method and computing equipment

Abstract: The invention discloses an image processing method and computing equipment. The image processing method comprises the following steps: acquiring a low-resolution image corresponding to an original image; inputting the low-resolution image and the original image as input images into an image processing machine learning model component; acquiring a high-resolution complete image with higher resolution and richer detail information compared with an original image, wherein the image processing machine learning model component is obtained by training a plurality of groups of image sets obtained inadvance, and the plurality of groups of image sets comprise a high-resolution complete image set, a high-resolution image set and a low-resolution image set.

Multi-component content prediction method and system in rare earth extraction process

Abstract: The invention provides a multi-component content prediction method under the condition of coexistence of rare earth ions with color characteristics and without color characteristics, and relates to the field of component content prediction in the rare earth extraction process. The method comprises: the problem that the component content is difficult to quickly and accurately detect exists in the rare earth extraction process; the GA-ELM-based rare earth extraction process multi-component content prediction method is provided for solving the problem that the original color feature-based singlerare earth element component content detection method is not applicable due to the fact that the image color features of a CePr/Nd mixed solution containing colorless Ce ions are greatly different from those of a Pr/Nd solution. The method comprises the following steps: firstly, searching H and S components with the maximum correlation with the component content in an HSI color space; secondly, establishing a multi-component content soft measurement model based on an extreme learning machine ELM by taking H and S component first moments as input; for the uncertainty of an initial weight and athreshold value of an ELM model, using a genetic algorithm GA to optimize model parameters, so that the precision of the optimized ELM model with the component content is higher.