一个波形导向的深度展开网络,用于单个图像的反射去除
概括
这项研究介绍了一种新的波形导向深层展开网络 (WDUNet),用于单个图像的反射去除. 通过分析图像频率,WDUNet有效地分离反射,显著提高了反射去除性能.
科学领域:
- 计算机视觉 计算机视觉
- 图像处理 图像处理
背景情况:
- 单图像反射去除 (SIRR) 是一个具有挑战性的计算机视觉任务.
- 深度学习方法已经推进了SIRR,但将类似的反射和传输内容分开仍然很困难.
研究的目的:
- 开发一种新的,可解释和可通用的SIRR方法.
- 为了利用频域分析来改善反射分离.
主要方法:
- 提出了一个使用离散波形变换 (DWT) 的波形导向深度展开网络 (WDUNet).
- 制定了一个基于优化的反射去除模型,并将其展开成一个神经网络.
- 集成的低频和高频参数估计模块 (LPEM/HPEM) 用于超参数优化.
主要成果:
- WDUNet有效地区分频域中的反射.
- 该网络在基准数据集上表现出高于最先进的方法的性能.
- 在客观指标和主观视觉质量方面取得了显著的改进.
结论:
- WDUNet为SIRR提供了增强的解释性和概括性.
- 频率意识的方法与深度展开相结合是非常有效的.
- 这种方法代表了单个图像反射去除技术的重大进步.
相关概念视频
Reflection of Waves
3.9K
When a wave travels from one medium to another, it gets reflected at the boundary of the second medium. A common example of this is when a person yells at a distance from a cliff and hears the echo of their voice. The sound waves (longitudinal waves) traveling in the air are reflected from the bounding cliff. Similarly, flipping one end of a string whose other end is tied to a wall causes a pulse (transverse wave) to travel through the string, which gets reflected upon reaching the wall. In...
3.9K
Deconvolution
262
Deconvolution, also known as inverse filtering, is the process of extracting the impulse response from known input and output signals. This technique is vital in scenarios where the system's characteristics are unknown, and they must be inferred from the observable signals.
Deconvolution involves several mathematical techniques to derive the impulse response. One common approach is polynomial division. In this method, the input and output sequences are treated as coefficients of...
Deconvolution involves several mathematical techniques to derive the impulse response. One common approach is polynomial division. In this method, the input and output sequences are treated as coefficients of...
262


