Compressed sensing reduces the complexity of signal acquisition but increases the complexity of signal
reconstruction.
Block-based intra-prediction algorithms have become popular for enhancing compression ratios.
Although parallel processing holds promise for reducing reconstruction time,
block interdependency poses a challenge.
This study introduces a reconstruction algorithm employing Zigzag ordering-based parallelism to address
these issues.
Experimental results show that the proposed algorithm accelerates the baseline algorithm by 3.26 to 7.13 times.
Top: The architecture of PGTformer.
Top: The architecture of PGTformer.
Quantitative comparison on VFHQ Blind setting.
This work is supported by JST, PRESTO Grant Number JPMJPR1757 Japan.
@inproceedings{10.1145/3447450.3447489,
author = {Xu, Jiayao and Peetakul, Jirayu and Li, Muchen and Zhou, Jinjia},
title = {High-speed Compressed Sensing Reconstruction using Zigzag Ordering based Parallel Processing},
year = {2021},
isbn = {9781450389075},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3447450.3447489},
doi = {10.1145/3447450.3447489},
abstract = {Compressed sensing (CS), as a signal processing technique, is often used to acquire and reconstruct a sparse signal. It can decrease the difficulty of acquiring signal while increase the difficulty of reconstructing the signal. Recently, block-based intra-prediction algorithms are widely used to further increase the compression ratio of images by using the information of neighboring blocks to predict the current block. However, it is hard to increase the speed by parallel processing due to the dependency among the blocks. Meanwhile, the reconstruction of compressed sensing images is time consuming. A reconstruction algorithm using Zigzag ordering-based parallelism is proposed in this paper to solve these problems. Besides, based on the feature of the chosen sensing matrix, a new method with higher efficiency for choosing the first candidate list in the reconstruction procedure was presented in this paper. The experimental results demonstrated that the proposed algorithm speedups the baseline algorithm for 3.26 to 7.13 times. And the quality of the reconstructed images is not changed. Thus, it is a promising solution for fast reconstruction of compressed images.},
booktitle = {Proceedings of the 2020 4th International Conference on Video and Image Processing},
pages = {247–255},
numpages = {9},
keywords = {Zigzag Scanning, Parallel Reconstruction, Block Compressed Sensing},
location = {Xi'an, China},
series = {ICVIP '20}
}