Learning Factorized Weight Matrix for Joint Filtering

Xiangyu Xu, Yongrui Ma, Wenxiu Sun
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:10587-10596, 2020.

Abstract

Joint filtering is a fundamental problem in computer vision with applications in many different areas. Most existing algorithms solve this problem with a weighted averaging process to aggregate input pixels. However, the weight matrix of this process is often empirically designed and not robust to complex input. In this work, we propose to learn the weight matrix for joint image filtering. This is a challenging problem, as directly learning a large weight matrix is computationally intractable. To address this issue, we introduce the correlation of deep features to approximate the aggregation weights. However, this strategy only uses inner product for the weight matrix estimation, which limits the performance of the proposed algorithm. Therefore, we further propose to learn a nonlinear function to predict sparse residuals of the feature correlation matrix. Note that the proposed method essentially factorizes the weight matrix into a low-rank and a sparse matrix and then learn both of them simultaneously with deep neural networks. Extensive experiments show that the proposed algorithm compares favorably against the state-of-the-art approaches on a wide variety of joint filtering tasks.

Cite this Paper


BibTeX
@InProceedings{pmlr-v119-xu20f, title = {Learning Factorized Weight Matrix for Joint Filtering}, author = {Xu, Xiangyu and Ma, Yongrui and Sun, Wenxiu}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {10587--10596}, year = {2020}, editor = {III, Hal Daumé and Singh, Aarti}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/xu20f/xu20f.pdf}, url = {http://proceedings.mlr.press/v119/xu20f.html}, abstract = {Joint filtering is a fundamental problem in computer vision with applications in many different areas. Most existing algorithms solve this problem with a weighted averaging process to aggregate input pixels. However, the weight matrix of this process is often empirically designed and not robust to complex input. In this work, we propose to learn the weight matrix for joint image filtering. This is a challenging problem, as directly learning a large weight matrix is computationally intractable. To address this issue, we introduce the correlation of deep features to approximate the aggregation weights. However, this strategy only uses inner product for the weight matrix estimation, which limits the performance of the proposed algorithm. Therefore, we further propose to learn a nonlinear function to predict sparse residuals of the feature correlation matrix. Note that the proposed method essentially factorizes the weight matrix into a low-rank and a sparse matrix and then learn both of them simultaneously with deep neural networks. Extensive experiments show that the proposed algorithm compares favorably against the state-of-the-art approaches on a wide variety of joint filtering tasks.} }
Endnote
%0 Conference Paper %T Learning Factorized Weight Matrix for Joint Filtering %A Xiangyu Xu %A Yongrui Ma %A Wenxiu Sun %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-xu20f %I PMLR %P 10587--10596 %U http://proceedings.mlr.press/v119/xu20f.html %V 119 %X Joint filtering is a fundamental problem in computer vision with applications in many different areas. Most existing algorithms solve this problem with a weighted averaging process to aggregate input pixels. However, the weight matrix of this process is often empirically designed and not robust to complex input. In this work, we propose to learn the weight matrix for joint image filtering. This is a challenging problem, as directly learning a large weight matrix is computationally intractable. To address this issue, we introduce the correlation of deep features to approximate the aggregation weights. However, this strategy only uses inner product for the weight matrix estimation, which limits the performance of the proposed algorithm. Therefore, we further propose to learn a nonlinear function to predict sparse residuals of the feature correlation matrix. Note that the proposed method essentially factorizes the weight matrix into a low-rank and a sparse matrix and then learn both of them simultaneously with deep neural networks. Extensive experiments show that the proposed algorithm compares favorably against the state-of-the-art approaches on a wide variety of joint filtering tasks.
APA
Xu, X., Ma, Y. & Sun, W.. (2020). Learning Factorized Weight Matrix for Joint Filtering. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:10587-10596 Available from http://proceedings.mlr.press/v119/xu20f.html.

Related Material