Efficient Medicinal Image Transmission and Resolution Enhancement via GAN

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sharma, Rishabh Kumar, Sharma, Mukund, Sharma, Pushkar, Aparjeeta, Jeetashree
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915026251546624
author Sharma, Rishabh Kumar
Sharma, Mukund
Sharma, Pushkar
Aparjeeta, Jeetashree
author_facet Sharma, Rishabh Kumar
Sharma, Mukund
Sharma, Pushkar
Aparjeeta, Jeetashree
contents While X-ray imaging is indispensable in medical diagnostics, it inherently carries with it those noises and limitations on resolution that mask the details necessary for diagnosis. B/W X-ray images require a careful balance between noise suppression and high-detail preservation to ensure clarity in soft-tissue structures and bone edges. While traditional methods, such as CNNs and early super-resolution models like ESRGAN, have enhanced image resolution, they often perform poorly regarding high-frequency detail preservation and noise control for B/W imaging. We are going to present one efficient approach that improves the quality of an image with the optimization of network transmission in the following paper. The pre-processing of X-ray images into low-resolution files by Real-ESRGAN, a version of ESRGAN elucidated and improved, helps reduce the server load and transmission bandwidth. Lower-resolution images are upscaled at the receiving end using Real-ESRGAN, fine-tuned for real-world image degradation. The model integrates Residual-in-Residual Dense Blocks with perceptual and adversarial loss functions for high-quality upscaled images with low noise. We further fine-tune Real-ESRGAN by adapting it to the specific B/W noise and contrast characteristics. This suppresses noise artifacts without compromising detail. The comparative evaluation conducted shows that our approach achieves superior noise reduction and detail clarity compared to state-of-the-art CNN-based and ESRGAN models, apart from reducing network bandwidth requirements. These benefits are confirmed both by quantitative metrics, including Peak Signal-to-Noise Ratio and Structural Similarity Index, and by qualitative assessments, which indicate the potential of Real-ESRGAN for diagnostic-quality X-ray imaging and for efficient medical data transmission.
format Preprint
id arxiv_https___arxiv_org_abs_2411_12833
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Efficient Medicinal Image Transmission and Resolution Enhancement via GAN
Sharma, Rishabh Kumar
Sharma, Mukund
Sharma, Pushkar
Aparjeeta, Jeetashree
Image and Video Processing
Artificial Intelligence
Computer Vision and Pattern Recognition
While X-ray imaging is indispensable in medical diagnostics, it inherently carries with it those noises and limitations on resolution that mask the details necessary for diagnosis. B/W X-ray images require a careful balance between noise suppression and high-detail preservation to ensure clarity in soft-tissue structures and bone edges. While traditional methods, such as CNNs and early super-resolution models like ESRGAN, have enhanced image resolution, they often perform poorly regarding high-frequency detail preservation and noise control for B/W imaging. We are going to present one efficient approach that improves the quality of an image with the optimization of network transmission in the following paper. The pre-processing of X-ray images into low-resolution files by Real-ESRGAN, a version of ESRGAN elucidated and improved, helps reduce the server load and transmission bandwidth. Lower-resolution images are upscaled at the receiving end using Real-ESRGAN, fine-tuned for real-world image degradation. The model integrates Residual-in-Residual Dense Blocks with perceptual and adversarial loss functions for high-quality upscaled images with low noise. We further fine-tune Real-ESRGAN by adapting it to the specific B/W noise and contrast characteristics. This suppresses noise artifacts without compromising detail. The comparative evaluation conducted shows that our approach achieves superior noise reduction and detail clarity compared to state-of-the-art CNN-based and ESRGAN models, apart from reducing network bandwidth requirements. These benefits are confirmed both by quantitative metrics, including Peak Signal-to-Noise Ratio and Structural Similarity Index, and by qualitative assessments, which indicate the potential of Real-ESRGAN for diagnostic-quality X-ray imaging and for efficient medical data transmission.
title Efficient Medicinal Image Transmission and Resolution Enhancement via GAN
topic Image and Video Processing
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2411.12833