An image reconstruction method and device based on a progressive convolution measurement network

Abstract

The invention provides an image reconstruction method and device based on a progressive convolution measurement network, and belongs to the technical field of image reconstruction. The method comprises the following steps: firstly, carrying out progressive convolution operation on an original image through a plurality of downsampling layers, and then generating a corresponding number of feature maps meeting a given measurement rate through a downsampling feature extraction layer; Carrying out progressive deconvolution operation on the feature map through a plurality of upper sampling layers corresponding to the plurality of lower sampling layers, and generating a preliminary reconstructed image consistent with the original image in size through an upper sampling feature extraction layer; And finally, carrying out quality optimization training on the preliminary reconstructed image by using a residual convolutional neural network to obtain a final feature map, and carrying out error processing on the final feature map and the original image to obtain a final optimized reconstructed image. According to the method, the image is sampled and reconstructed end to end, the reconstructionspeed is high, the block effect in the reconstructed image is eliminated especially at an extremely low measurement rate, and the image quality is obviously improved.

Description

technical field [0001] The invention relates to the technical field of image reconstruction, in particular to an image reconstruction method and device based on a progressive convolution measurement network. Background technique [0002] Compressed sensing (Compressive Sensing, CS), as a new type of data compression technology, has received widespread attention. Using fewer measurements than required by the Nyquist sampling theorem, CS theory demonstrates that when a signal exhibits sparsity under certain conditions, it is very likely to be fully reconstructed. Mathematically, the goal of CS reconstruction is to measure Y=ΦX∈R from CS M×1 The original signal X ∈ R is deduced from N×1 . Here, Φ∈R M×N is a linear random matrix. Random Gaussian matrices are commonly used as measurement matrices, because we must ensure that the basis of the sparse domain of the matrix is ​​completely inconsistent with the measurement. Because M<<N, the inverse problem is usually ill-...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to provide an image reconstruction method based on a progressive convolution measurement network to solve the problem that the existing image reconstruction method in the above-mentioned background technology can only test a fixed-size image and the reconstructed image has block effects technical problem

Images