A threshold-based coefficients cutting method for compressive imaging

作者：Quan Lei^*; Xiao Song^*; Du Jianchao^*; Zhou Janli^*

来源：7th International Conference on Intelligent Information Hiding and Multimedia Signal Processing, IIHMSP 2011, 2011-10-14 to 2011-10-16.

DOI：10.1109/IIHMSP.2011.96

• Best approximations
• Coding scheme
• coefficients-cutting
• Compressive imaging
• compressive sensing
• Cutting algorithm
• Joint source
• Linear measurements
• Natural images
• Original images
• Reconstruction error
• Signal vectors
• Sparse representation
• Sparse signals
• Total variation
• Transform domain

Compressive sensing is a new technique in signal processing which can recover a sparse signal vector via a much smaller of non-adaptive, linear measurements than the dimension of the signal vector. In this paper, we applied compressive sensing to a joint source compression-channel coding scheme. With analysis of the reconstruction error of the sparse representation of natural image signals, we argue that recovering the large coefficients exactly as many as possible in the transform domain would obtain higher PSNR than reconstructing the best approximation of the original images directly. Then a threshold-based coefficient cutting method was proposed in order to guarantee the accurate recovery of the large coefficients. The experimental results show that with the proposed coefficient cutting algorithm, the reconstructed quality of image could be greatly improved and matches or exceeds that of the popular but computationally expensive Minimizing Total Variation algorithm.

• 出版日期2011
• 单位西安电子科技大学