A Distributed Compressed Sensing for Images Based on Block Measurements Data Fusion

Huaixin Chen, Jie LiuJournal of Software Engineering and Applications 2012 · DOI

Abstract

Compressed sensing (CS) is a new technique for simultaneous data sampling and compression. In this paper, we propose a novel method called distributed compressed sensing for image using block measurements data fusion. Firstly, original image is divided into small blocks and each block is sampled independently using the same measurement operator, to obtain the smaller encoded sparser coefficients and stored measurements matrix and its vectors. Secondly, original image is reconstructed using the block measurements fusion and recovery transform. Finally, several numerical experiments demonstrate that our method has a much lower data storage and calculation cost as well as high quality of reconstruction when compared with other existing schemes. We believe it is of great practical potentials in the network communication as well as pattern recognition domain.

Introduction

Compressed Sensing (CS) has emerged as a powerful framework for capturing signals using significantly fewer samples than required by the Nyquist-Shannon theorem. This efficiency is particularly valuable for low-power sensors and bandwidth-constrained networks. However, applying CS to high-resolution images presents significant computational challenges.

To mitigate these, researchers often employ Block-based CS (BCS), where an image is partitioned into small sub-blocks for independent sampling. The traditional reconstruction method, known as Sub-Image Jointing (SIJ-BCS), reconstructs each block individually before stitching them together. This approach suffers from two primary drawbacks: high computational overhead due to managing numerous measurement matrices and the presence of "blocking artifacts" caused by ignoring spatial correlations between neighboring blocks.

Figure 1: High-level overview of a distributed compressed sensing system architecture.

Methodology: Block Measurements Data Fusion (BMF-DCS)

This paper introduces Block Measurements Data Fusion (BMF-DCS), a "fusion-first" architecture designed to address the limitations of traditional BCS. The proposed pipeline shifts the complexity away from individual block reconstruction toward a unified global reconstruction process.

The BMF-DCS Pipeline

  1. Encoder (Sampling): The original image is divided into sub-blocks. Each block is sampled independently using a random Gaussian measurement operator to generate sparse coefficients.
  2. Data Fusion: The critical innovation occurs at the decoder. Instead of processing blocks sequentially, the decoder mathematically fuses the sub-block measurement matrices and vectors into a single, synthesized global system. This process leverages the linearity of the measurement process ($y = Ax$) to create a comprehensive representation of the image before reconstruction begins.
  3. Decoder (Reconstruction): The entire image is reconstructed from the fused data using the Orthogonal Matching Pursuit (OMP) algorithm, with a Discrete Cosine Transform (DCT) matrix serving as the sparsity basis.
Figure 6: The proposed BMF-DCS architecture, highlighting the block measurements fusion step in the decoder.

Experimental Results

The BMF-DCS method was evaluated against standard grayscale test images (Dock, Mountain, and Cameraman) at a measurement ratio of 0.6. The results demonstrate significant improvements over traditional SIJ-BCS across three key metrics:

  • Computational Efficiency: The Time of Decoding (TOD) was drastically reduced. For the "Dock" image, processing time fell from 4.336 seconds to just 0.031 seconds—a nearly 140x speedup.
  • Storage Requirements: The Number of Data stored in the Decoder (NDD) was reduced by approximately 40%, dropping from 65,536 to 39,321 units. This makes the method highly suitable for hardware with limited memory.
  • Image Quality: The Peak Signal-to-Noise Ratio (PSNR) improved across all test cases. Notably, the "Mountain" image saw a 4.5 dB increase in quality. By reconstructing the image as a whole, BMF-DCS naturally accounts for inter-block correlations, effectively eliminating blocking artifacts.

Conclusion

BMF-DCS represents a significant advancement in distributed compressed sensing. By fusing measurements before reconstruction, the method achieves superior image quality while simultaneously reducing storage and computational costs. Future work will focus on scaling this approach to higher-resolution images and evaluating its robustness against transmission noise in real-world sensor networks.

Summary updated

Abstract

Compressed sensing (CS) is a new technique for simultaneous data sampling and compression. In this paper, we propose a novel method called distributed compressed sensing for image using block measurements data fusion. Firstly, original image is divided into small blocks and each block is sampled independently using the same measurement operator, to obtain the smaller encoded sparser coefficients and stored measurements matrix and its vectors. Secondly, original image is reconstructed using the block measurements fusion and recovery transform. Finally, several numerical experiments demonstrate that our method has a much lower data storage and calculation cost as well as high quality of reconstruction when compared with other existing schemes. We believe it is of great practical potentials in the network communication as well as pattern recognition domain.
BibTeX
@article{chen2012distributed,
  author = {Huaixin Chen and Jie Liu},
  title = {{A Distributed Compressed Sensing for Images Based on Block Measurements Data Fusion}},
  journal = {Journal of Software Engineering and Applications},
  year = {2012},
  doi = {10.4236/jsea.2012.512b026},
}
Figure