A Method for Multi-image Compression and Reconstruction Based on Compressed Sensing and Orthogonal Modulation

Abstract

The invention discloses a method for realizing multi-image compression and reconstruction based on compressed sensing and orthogonal modulation, which belongs to the field of optical single-pixel imaging and can solve the problems caused by compressed sampling and reconstruction of huge images in traditional single-pixel cameras. Even if it only needs to reconstruct local block images, it needs a lot of data processing. The present invention mainly utilizes the orthogonal modulation technology to realize the block reconstruction of the image or the partial reconstruction in multiple images, and its steps can be summarized as follows: first, the original image is evenly divided into blocks, and the perceptual matrix corresponding to each block image multiplied by the corresponding orthogonal basis matrix to obtain a new random matrix containing orthogonal information, which is equivalent to the sensing matrix in traditional compressed sensing; then use the principle of compressed sensing to collect data; in image reconstruction, just use The transpose of the corresponding orthogonal basis matrix is ​​multiplied by the collected total data, and the corresponding block image data can be extracted, and then the reconstruction operation can be performed. This method can restore only the required part of the image data, thus reducing the burden of data processing.

Description

technical field [0001] The invention belongs to the field of optical single-pixel imaging, in particular to a dynamic illumination single-pixel imaging method. Background technique [0002] The traditional Nyquist sampling law tells us that in order to accurately reconstruct the original signal, it must be sampled at twice the highest frequency of the signal. However, this method produces huge redundant data, only a small amount of data is used, and the bandwidth utilization rate of the signal is not high. Therefore, how to represent signals with fewer bits to reduce the burden of storage, transmission and processing is a problem we must solve. Donoho proposed the theory of compressed sensing (Donoho DL. Compressed sensing [J]. IEEE Transaction on Information Theory. 2006, 52(4): 1289-1306.), which removes the redundancy by means of compressed sensing (or direct sensing). The information after the remaining data realizes the process of collecting and compressing at the sam...

AI Technical Summary

Problems solved by technology

For multiple images, it is necessary to repeat the compression acquisition process of a single image multiple times and record a large amount of data; or for a huge image composed of several block sub-images, when only part of the block sub-image information needs to be reconstructed , but has to restore and reconstruct the entire huge image, resulting in some unnecessary data reconstruction, greatly increasing the amount of information processing, and wasting computational overhead

Images