Minimizing Acquisition Maximizing Inference -- A demonstration on print error detection

TL;DR

This paper explores a method to detect features in images with minimal measurements, balancing between data acquisition and privacy, demonstrated through print error detection.

Contribution

It introduces a novel approach combining compressed sensing and feature inference to detect image features with fewer measurements for privacy preservation.

Findings

Effective feature detection with fewer measurements

Potential for privacy-preserving image analysis

Demonstrated on print error detection task

Abstract

Is it possible to detect a feature in an image without ever looking at it? Images are known to have sparser representation in Wavelets and other similar transforms. Compressed Sensing is a technique which proposes simultaneous acquisition and compression of any signal by taking very few random linear measurements (M). The quality of reconstruction directly relates with M, which should be above a certain threshold for a reliable recovery. Since these measurements can non-adaptively reconstruct the signal to a faithful extent using purely analytical methods like Basis Pursuit, Matching Pursuit, Iterative thresholding, etc., we can be assured that these compressed samples contain enough information about any relevant macro-level feature contained in the (image) signal. Thus if we choose to deliberately acquire an even lower number of measurements - in order to thwart the possibility of a comprehensible reconstruction, but high enough to infer whether a relevant feature exists in an image - we can achieve accurate image classification while preserving its privacy. Through the print error detection problem, it is demonstrated that such a novel system can be implemented in practise.