Multi-scale Deep Convolutional Neural Networks for Microscopic Image Super-resolution

DOI

10.2991/978-94-6463-602-4_6How to use a DOI?

Keywords

Super-resolution microscopy; Single image super-resolution (SISR); Biomedical imaging; Depthwise Separable Convolution

Abstract

Deep convolutional neural networks (CNNs) have recently shown remarkable success in single image super-resolution (SISR), particularly in medical image super-resolution for microscopy. However, microscopy image reconstruction remains a challenging task through conventional approaches, which often require high hardware costs and yield unsatisfactory results. We propose a new multi-scale deep CNN architecture tailored to SISR for low-resolution (LR) microscopic images. To tackle the challenges of training deep CNNs, we utilize a residual learning approach, explicitly supervising the residuals using the disparity between high-resolution (HR) and LR images. The sum up of the recovered details to the LR image results in the reconstruction of the HR image. In addition, we employ gradient clipping to prevent gradient explosions that can occur with high learning rates. Furthermore, the choice of Depthwise separable convolution in our paper is to justified by its ability to reduce computational complexity and less memory usage while maintaining high accuracy. In contrast to current deep CNN-based SISR methods for natural images, where LR images are received by subsampling and blurring HR images, we evaluate our approach using lower objective lenses and thin smear blood samples. HR images captured with higher objective lenses are used as a benchmark to compare the performance. Extensive evaluations confirm that Multi-Scale Deep Convolutional Neural Networks for Microscopic Image Super-resolution (MDCM) outperform other methods. The proposed MDCM method addresses the critical need for accurate and fast reconstruction algorithms to improve temporal resolution in high-density super-resolution microscopy, particularly for live-cell imaging.

Copyright

© 2024 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Wazir Muhammad
AU  - Nazia Ejaz
AU  - Ayaz Hussain
AU  - Jalal Shah
AU  - Sohrab Khan
AU  - Inam Ul Ahad
PY  - 2024
DA  - 2024/12/24
TI  - Multi-scale Deep Convolutional Neural Networks for Microscopic Image Super-resolution
BT  - Proceedings of the 4th International Conference on Key Enabling Technologies (KEYTECH 2024)
PB  - Atlantis Press
SP  - 41
EP  - 49
SN  - 2589-4943
UR  - https://doi.org/10.2991/978-94-6463-602-4_6
DO  - 10.2991/978-94-6463-602-4_6
ID  - Muhammad2024
ER  -