Characterization of a compressive imaging system using laboratory and natural light scenes

Area of Science:

• Optics and Photonics
• Image Acquisition Technology

Background:

• Compressive imaging (CI) provides a novel approach to optical sensing, potentially enhancing size, weight, and performance (SWaP).
• CI systems acquire images through spatially filtered intensity measurements, with data requirements varying by desired image quality.

Purpose of the Study:

• To conduct the first systematic performance comparison between a compressive imaging system and a conventional focal plane imager.
• To evaluate CI performance across various scene types including binary, grayscale, and natural light (color and infrared).

Main Methods:

• Generated 1024x1024 images using compressive imaging with digital (Hadamard), grayscale (discrete cosine transform), and random (Noiselet) basis sets.
• Acquired images across a range of measurement percentages (0.1%-100%).
• Compared compressive images against conventionally acquired images, each using only 1% of full sampling.

Main Results:

• The Hadamard Transform basis set exhibited the best performance among the tested CI methods.
• Compressive images generated using the Hadamard Transform showed aesthetic quality comparable to conventional images.
• Hadamard Transform-based CI yielded slightly higher spatial resolution than conventional imaging at 1% sampling.

Conclusions:

• Compressive imaging, particularly with the Hadamard Transform, is a viable and effective method for optical sensing.
• CI systems can achieve high-quality image results with significantly reduced data acquisition compared to traditional methods.
• The Hadamard Transform offers a promising basis set for optimizing compressive imaging performance for various applications.