Speckle Reduction with Trained Nonlinear Diffusion Filtering

Wensen Feng, Yunjin Chen

Publikation: Beitrag in einer FachzeitschriftArtikelForschungBegutachtung

Abstract

Speckle reduction is a prerequisite for many image processing tasks in synthetic aperture radar images, as well as all coherent images. In recent years, predominant state-of-the-art approaches for despeckling are usually based on nonlocal methods which mainly concentrate on achieving utmost image restoration quality, with relatively low computational efficiency. Therefore, in this study we aim to propose an efficient despeckling model with both high computational efficiency and high recovery quality. To this end, we exploit a newly developed trainable nonlinear reaction diffusion (TNRD) framework which has proven a simple and effective model for various image restoration problems. In the original TNRD applications, the diffusion network is usually derived based on the direct gradient descent scheme. However, this approach will encounter some problem for the task of multiplicative noise reduction exploited in this study. To solve this problem, we employed a new architecture derived from the proximal gradient descent method. Taking into account the speckle noise statistics, the diffusion process for the despeckling task is derived. We then retrain all the model parameters in the presence of speckle noise. Finally, optimized nonlinear diffusion filtering models are obtained, which are specialized for despeckling with various noise levels. Experimental results substantiate that the trained filtering models provide comparable or even better results than state-of-the-art nonlocal approaches. Meanwhile, our proposed model merely contains convolution of linear filters with an image, which offers high-level parallelism on GPUs. As a consequence, for images of size 512 × 512 , our GPU implementation takes less than 0.1 s to produce state-of-the-art despeckling performance.

Originalspracheenglisch
Seiten (von - bis)162-178
Seitenumfang17
FachzeitschriftJournal of Mathematical Imaging and Vision
Jahrgang58
Ausgabenummer1
DOIs
PublikationsstatusVeröffentlicht - 1 Mai 2017

Fingerprint

Nonlinear Diffusion
Speckle
Filtering
Speckle Noise
Image Restoration
descent
Reaction-diffusion
Computational efficiency
Image reconstruction
restoration
Computational Efficiency
Model
linear filters
Gradient Descent Method
gradients
Linear Filter
Multiplicative Noise
Gradient Descent
Noise Reduction
Synthetic Aperture

Schlagwörter

    ASJC Scopus subject areas

    • !!Statistics and Probability
    • !!Modelling and Simulation
    • !!Condensed Matter Physics
    • !!Computer Vision and Pattern Recognition
    • !!Geometry and Topology
    • Angewandte Mathematik

    Dies zitieren

    Speckle Reduction with Trained Nonlinear Diffusion Filtering. / Feng, Wensen; Chen, Yunjin.

    in: Journal of Mathematical Imaging and Vision, Jahrgang 58, Nr. 1, 01.05.2017, S. 162-178.

    Publikation: Beitrag in einer FachzeitschriftArtikelForschungBegutachtung

    Feng, Wensen ; Chen, Yunjin. / Speckle Reduction with Trained Nonlinear Diffusion Filtering. in: Journal of Mathematical Imaging and Vision. 2017 ; Jahrgang 58, Nr. 1. S. 162-178.
    @article{0db6ec3807c34d5fbf020c04d5300768,
    title = "Speckle Reduction with Trained Nonlinear Diffusion Filtering",
    abstract = "Speckle reduction is a prerequisite for many image processing tasks in synthetic aperture radar images, as well as all coherent images. In recent years, predominant state-of-the-art approaches for despeckling are usually based on nonlocal methods which mainly concentrate on achieving utmost image restoration quality, with relatively low computational efficiency. Therefore, in this study we aim to propose an efficient despeckling model with both high computational efficiency and high recovery quality. To this end, we exploit a newly developed trainable nonlinear reaction diffusion (TNRD) framework which has proven a simple and effective model for various image restoration problems. In the original TNRD applications, the diffusion network is usually derived based on the direct gradient descent scheme. However, this approach will encounter some problem for the task of multiplicative noise reduction exploited in this study. To solve this problem, we employed a new architecture derived from the proximal gradient descent method. Taking into account the speckle noise statistics, the diffusion process for the despeckling task is derived. We then retrain all the model parameters in the presence of speckle noise. Finally, optimized nonlinear diffusion filtering models are obtained, which are specialized for despeckling with various noise levels. Experimental results substantiate that the trained filtering models provide comparable or even better results than state-of-the-art nonlocal approaches. Meanwhile, our proposed model merely contains convolution of linear filters with an image, which offers high-level parallelism on GPUs. As a consequence, for images of size 512 × 512 , our GPU implementation takes less than 0.1 s to produce state-of-the-art despeckling performance.",
    keywords = "Convolutional neural networks, Despeckling, Optimized nonlinear reaction diffusion model, Trainable activation function",
    author = "Wensen Feng and Yunjin Chen",
    year = "2017",
    month = "5",
    day = "1",
    doi = "10.1007/s10851-016-0697-x",
    language = "English",
    volume = "58",
    pages = "162--178",
    journal = "Journal of Mathematical Imaging and Vision",
    issn = "0924-9907",
    publisher = "Springer Science+Business Media B.V",
    number = "1",

    }

    TY - JOUR

    T1 - Speckle Reduction with Trained Nonlinear Diffusion Filtering

    AU - Feng, Wensen

    AU - Chen, Yunjin

    PY - 2017/5/1

    Y1 - 2017/5/1

    N2 - Speckle reduction is a prerequisite for many image processing tasks in synthetic aperture radar images, as well as all coherent images. In recent years, predominant state-of-the-art approaches for despeckling are usually based on nonlocal methods which mainly concentrate on achieving utmost image restoration quality, with relatively low computational efficiency. Therefore, in this study we aim to propose an efficient despeckling model with both high computational efficiency and high recovery quality. To this end, we exploit a newly developed trainable nonlinear reaction diffusion (TNRD) framework which has proven a simple and effective model for various image restoration problems. In the original TNRD applications, the diffusion network is usually derived based on the direct gradient descent scheme. However, this approach will encounter some problem for the task of multiplicative noise reduction exploited in this study. To solve this problem, we employed a new architecture derived from the proximal gradient descent method. Taking into account the speckle noise statistics, the diffusion process for the despeckling task is derived. We then retrain all the model parameters in the presence of speckle noise. Finally, optimized nonlinear diffusion filtering models are obtained, which are specialized for despeckling with various noise levels. Experimental results substantiate that the trained filtering models provide comparable or even better results than state-of-the-art nonlocal approaches. Meanwhile, our proposed model merely contains convolution of linear filters with an image, which offers high-level parallelism on GPUs. As a consequence, for images of size 512 × 512 , our GPU implementation takes less than 0.1 s to produce state-of-the-art despeckling performance.

    AB - Speckle reduction is a prerequisite for many image processing tasks in synthetic aperture radar images, as well as all coherent images. In recent years, predominant state-of-the-art approaches for despeckling are usually based on nonlocal methods which mainly concentrate on achieving utmost image restoration quality, with relatively low computational efficiency. Therefore, in this study we aim to propose an efficient despeckling model with both high computational efficiency and high recovery quality. To this end, we exploit a newly developed trainable nonlinear reaction diffusion (TNRD) framework which has proven a simple and effective model for various image restoration problems. In the original TNRD applications, the diffusion network is usually derived based on the direct gradient descent scheme. However, this approach will encounter some problem for the task of multiplicative noise reduction exploited in this study. To solve this problem, we employed a new architecture derived from the proximal gradient descent method. Taking into account the speckle noise statistics, the diffusion process for the despeckling task is derived. We then retrain all the model parameters in the presence of speckle noise. Finally, optimized nonlinear diffusion filtering models are obtained, which are specialized for despeckling with various noise levels. Experimental results substantiate that the trained filtering models provide comparable or even better results than state-of-the-art nonlocal approaches. Meanwhile, our proposed model merely contains convolution of linear filters with an image, which offers high-level parallelism on GPUs. As a consequence, for images of size 512 × 512 , our GPU implementation takes less than 0.1 s to produce state-of-the-art despeckling performance.

    KW - Convolutional neural networks

    KW - Despeckling

    KW - Optimized nonlinear reaction diffusion model

    KW - Trainable activation function

    UR - http://www.scopus.com/inward/record.url?scp=85007442768&partnerID=8YFLogxK

    U2 - 10.1007/s10851-016-0697-x

    DO - 10.1007/s10851-016-0697-x

    M3 - Article

    VL - 58

    SP - 162

    EP - 178

    JO - Journal of Mathematical Imaging and Vision

    JF - Journal of Mathematical Imaging and Vision

    SN - 0924-9907

    IS - 1

    ER -