Fast End-to-End Trainable Guided Filter

By Huikai Wu, Shuai Zheng, Junge Zhang, Kaiqi Huang. 2018

266 citations recorded by the source corpus.

Image processing and pixel-wise dense prediction have been advanced by harnessing the capabilities of deep learning. One central issue of deep learning is the limited capacity to handle joint upsampling. We present a deep learning building block for joint upsampling, namely guided filtering layer. This layer aims at efficiently generating the high-resolution output given the corresponding low-resolution one and a high-resolution guidance map. The proposed layer is composed of a guided filter, which is reformulated as a fully differentiable block. To this end, we show that a guided filter can be expressed as a group of spatial varying linear transformation matrices. This layer could be integrated with the convolutional neural networks (CNNs) and jointly optimized through end-to-end training. To further take advantage of end-to-end training, we plug in a trainable transformation function that generates task-specific guidance maps. By integrating the CNNs and the proposed layer, we form deep guided filtering networks. The proposed networks are evaluated on five advanced image processing tasks. Experiments on MIT-Adobe FiveK Dataset demonstrate that the proposed approach runs 10-100Ã— faster and achieves the state-of-the-art performance. We also show that the proposed guided filtering layer helps to improve the performance of multiple pixel-wise dense prediction tasks. The code is available at https://github.com/wuhuikai/DeepGuidedFilter.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Getting to 99% Accuracy in Interactive Segmentation (2020, methodology): “We further incorporate a Guided Filter layer [19] to refine the results and produce precise transparency masks.”
Heavy Rain Image Restoration: Integrating Physics Model and Conditional Adversarial Learning (2019, methodology): “Our decomposition is adopted from [37], where we create a decomposition CNN layer that is differentiable during training (details shown in Fig.”
Low-Light Image Enhancement via a Deep Hybrid Network (2019, background): “For computer vision tasks, the number of affine coefficients is usually very large, which becomes the performance and speed bottlenecks [25].”
Learning Guided Convolutional Network for Depth Completion (2019, result): “In [43], Wu et al. propose a guided ﬁltering layer to perform joint upsampling, which is close to our work.”
Deep Interactive Thin Object Selection (2021, methodology): “Speciﬁcally, we insert a deep guided ﬁltering layer (DGF) [46] at the end of M2 to recover the details of elon-gated thin structures when upsampling back to its original resolution.”
DeepLandscape: Adversarial Modeling of Landscape Video (2020, methodology): “The static parts are obtained by running the guided ﬁlter [36] on the super-resolved frames while using the input high-res image as a guide.”
ALL Snow Removed: Single Image Desnowing Algorithm Using Hierarchical Dual-tree Complex Wavelet Representation and Contradict Channel Loss (2021, methodology): “In Table 3 and Figure 8, we adopt the vanilla convolution with kernel size 3 × 3 (VC), the multi-convolution (MC) [41], image-smoothing based high-low frequency decomposition (HLD) [39, 42], and the DWT [24].”
Pixel-Adaptive Convolutional Neural Networks (2019, background): “[45] and Wang et al .”
FastFCN: Rethinking Dilated Convolution in the Backbone for Semantic Segmentation (2019, methodology): “[31] proposes an end-to-end trainable guided ﬁltering module, which upsamples a low-resolution image conditionally.”