一个高度强大的编码器解码器网络,具有多级特征增强和注意力门,用于减少数字图像中的混合高斯式和盐和胡噪声
Milan Tripathi1, Waree Kongprawechnon1, Toshiaki Kondo1
1School of Information, Computer and Communication Technology, Sirindhorn International Institute of Technology, Thammasat University, Pathum Thani 12120, Thailand.
Journal of imaging
|February 25, 2025
概括
一个新的高强度编码解码器网络 (HREDN) 有效地从图像中去除混合噪声. 这种先进的图像无色化方法保留了关键细节,并在各种应用中优于现有的技术.
科学领域:
- 计算机视觉 计算机视觉
- 图像处理 图像处理
- 人工智能的人工智能
背景情况:
- 由于环境因素和技术限制造成的图像扭曲需要有效地消除.
- 混合噪声,包括盐和胡和高斯噪声,在数字图像处理中构成重大挑战.
研究的目的:
- 引入一种新的高强度编码解码器网络 (HREDN),以实现有效的图像消噪.
- 为了提高在剥光过程中图像细节和结构特征的保存.
主要方法:
- 在编码器中,HREDN包含了一个多级特征增强块,用于全面的特征捕获.
- 跳过连接与集成的注意力门被用来传输相关的功能和减轻信息丢失.
- 网络架构旨在处理复杂的噪声模式并保持结构完整性.
主要成果:
- 在消除混合盐和胡和高斯噪声方面,HREDN表现出卓越的性能.
- 拟议的方法在保留边缘细节和结构特征的图像中表现出色.
- 定量和定性评估表明HREDN的表现优于最先进的宣传技术.
结论:
- HREDN提供了一个强大的解决方案,用于在不同的领域,包括面部,医疗和遥感图像 denoising.
- 该模型有效地处理复杂的噪音场景和不同的图像分辨率.
- 注意性跳过连接是保存基本图像信息的关键.
相关概念视频
Downsampling
123
When considering a sampled sequence with zero values between sampling instants, one can replace it by taking every N-th value of the sequence. At these integer multiples of N, the original and sampled sequences coincide. This process, known as decimation, involves extracting every N-th sample from a sequence, thereby creating a more efficient sequence.
The Fourier transform of the decimated sequence reveals a combination of scaled and shifted versions of the original spectrum. This...
The Fourier transform of the decimated sequence reveals a combination of scaled and shifted versions of the original spectrum. This...
123
Upsampling
189
Managing signal sampling rates is essential in digital signal processing to maintain signal integrity. A decimated signal, characterized by a reduced frequency range due to its lower sampling rate, can be upsampled by inserting zeros between each sample. This upsampling process expands the original spectrum and introduces repeated spectral replicas at intervals dictated by the new Nyquist frequency. To refine this zero-inserted sequence, it is passed through a lowpass filter with a cutoff...
189


