Binned progressive quantization for compressive sensing.

Area of Science:

• Signal Processing
• Information Theory
• Data Compression

Background:

• Compressive Sensing (CS) offers architectural advantages for signal acquisition and communication in resource-limited scenarios.
• However, CS typically exhibits suboptimal rate-distortion performance compared to conventional methods.
• Existing CS encoders are signal-independent and computationally simple, shifting complexity to the decoder.

Purpose of the Study:

• To propose a novel coding technique that enhances the compression performance of Compressive Sensing (CS).
• To maintain the encoder simplicity and universality characteristic of current CS designs.
• To enable the CS decoder to leverage previously overlooked correlations within CS measurements.

Main Methods:

• Development of a progressive fixed-rate scalar quantization scheme with binning.
• Integration of this scheme into the CS decoding process.
• Experimental evaluation using test images to assess performance.

Main Results:

• The proposed technique partially rectifies the poor compression performance of CS.
• The CS decoder can exploit hidden correlations between CS measurements.
• On certain test images, the new CS coding technique achieves performance comparable to or exceeding JPEG.

Conclusions:

• The novel CS coding technique offers improved rate-distortion performance while preserving encoder simplicity.
• The method effectively utilizes inter-measurement correlations, a previously unaddressed aspect in CS.
• This advancement presents a viable alternative for efficient signal compression in CS systems.