Adaptive Rate Block Compressive Sensing Based on Statistical Characteristics Estimation

Area of Science:

• Signal Processing
• Computer Vision
• Video Compression

Background:

• Adaptive rate compressive sensing (ARCS) is challenging for devices with limited computing power, memory, and energy.
• Existing methods struggle when the original signal's sparsity is unknown to the sampling device.

Purpose of the Study:

• To propose a novel blocked ARCS method for surveillance videos that addresses device limitations.
• To enable efficient video sampling in resource-constrained environments like wireless video sensor networks (WVSN) and single pixel cameras (SPC).

Main Methods:

• Estimating original signal statistical characteristics by observing compressive sensing (CS) measurement results.
• Reasonably estimating signal sparsity using these statistical characteristics.
• Dividing video blocks into more classes with higher accuracy for adaptive sampling.

Main Results:

• The proposed method demonstrates low computational complexity, small memory footprint, and low power consumption.
• Effective adaptation to sparsity changes, appropriate sampling rate allocation per block, and reduced overall sampling rates.
• Improved reconstructed image quality, significantly lower sampling process computation, and accelerated sampling speed.

Conclusions:

• The blocked ARCS method is suitable for resource-limited applications like WVSN and SPC.
• The method effectively balances sampling rate reduction and reconstructed image quality.
• Outperforms previous state-of-the-art methods in overall performance.