Direct Image Classification from Fourier Ptychographic Microscopy Measurements without Reconstruction

2025-05-08

AI Summary

Fourier ptychographic microscopy (FPM) suffers from high computational overhead in high-resolution image reconstruction, hindering real-time classification. Method: We propose a measurement-domain end-to-end classification framework that bypasses conventional image reconstruction. Specifically, we design a learnable measurement reuse mechanism to directly model bandwidth-limited, multi-angle, low-resolution FPM measurements, integrated with a lightweight CNN for joint optimization. Contribution/Results: This is the first work to achieve direct classification in the FPM measurement domain, drastically reducing data dimensionality and acquisition time. Our method improves classification accuracy by 12% over single-bandlimited-image baselines and accelerates inference by an order of magnitude compared to reconstruction-then-classification pipelines. The framework establishes a new task-driven paradigm for efficient perception in computational imaging systems.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Multi-modal Vision
• Search and Optimization: Learning to Search

Application Category

• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Web Mining and Content Analysis: Web measurements

Abstract

The computational imaging technique of Fourier Ptychographic Microscopy (FPM) enables high-resolution imaging with a wide field of view and can serve as an extremely valuable tool, e.g. in the classification of cells in medical applications. However, reconstructing a high-resolution image from tens or even hundreds of measurements is computationally expensive, particularly for a wide field of view. Therefore, in this paper, we investigate the idea of classifying the image content in the FPM measurements directly without performing a reconstruction step first. We show that Convolutional Neural Networks (CNN) can extract meaningful information from measurement sequences, significantly outperforming the classification on a single band-limited image (up to 12 %) while being significantly more efficient than a reconstruction of a high-resolution image. Furthermore, we demonstrate that a learned multiplexing of several raw measurements allows maintaining the classification accuracy while reducing the amount of data (and consequently also the acquisition time) significantly.

Problem

• Classify images directly from FPM measurements without reconstruction
• Reduce computational cost of high-resolution image reconstruction
• Maintain accuracy while minimizing data acquisition time

Innovation

• Classify FPM measurements directly without reconstruction
• Use CNNs to extract info from measurement sequences
• Learned multiplexing reduces data and maintains accuracy