Deep Morphological Filter Networks For Gaussian Denoising

Fujisaki, Hikaru; Nakashizuka, Makoto

doi:10.1109/icip40778.2020.9190657

Cited by 7 publications

(10 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Every neuron operates in a tropical semiring, i.e., performs addition/subtraction and min/max operations. The recent research demonstrates the applicability of morphological neural networks for Gaussian denoising [18,19]. However, in practical applications, such models still have limited usage due to unacceptable quality for many standard tasks.…”

Section: Related Workmentioning

confidence: 99%

Quantization method for bipolar morphological neural networks

Limonova,

Zingerenko,

Nikolaev

et al. 2024

Sixteenth International Conference on Machine Vision (ICMV 2023)

View full text Add to dashboard Cite

In the paper, we present a quantization method for bipolar morphological neural networks. Bipolar morphological neural networks use only addition, subtraction, and maximum operations inside the neuron and exponent and logarithm as activation functions of the layers. These operations allow fast and compact gate implementation for FPGA and ASIC, which makes these networks a promising solution for embedded devices. Quantization allows us to reach an additional increase in computational efficiency and reduce the complexity of hardware implementation by using integer values of low bitwidth for computations. We propose an 8-bit quantization scheme based on integer maximum, addition, and lookup tables for non-linear functions and experimentally demonstrate that basic models for image classification can be quantized without noticeable accuracy loss. More advanced models still provide high recognition accuracy but would benefit from further fine-tuning.

show abstract

Section: Related Workmentioning

confidence: 99%

Quantization method for bipolar morphological neural networks

Limonova,

Zingerenko,

Nikolaev

et al. 2024

Sixteenth International Conference on Machine Vision (ICMV 2023)

View full text Add to dashboard Cite

show abstract

“…When comparing with (11), additional K multiplications and parameters are included. The additional parameters will contribute the improve the performance of the applications.…”

Section: Linear Combination Of Morphological Laplacianmentioning

confidence: 99%

“…In this subsection, the deep networks are developed by the morphological Laplacians. By substituting the Laplacian (11) defined by MoD (9) and MoE (10) to the iteration (3), we obtain…”

Section: Unrolling Of Diffusion Processmentioning

confidence: 99%

See 1 more Smart Citation

Deep Unrolling of Diffusion Process with Morphological Laplacian and its Implementation with SIMD Instructions

Okada

Nakashizuka

2022

2022 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

This paper presents a deep network based on unrolling of a diffusion process with morphological Laplacian. The diffusion process is an iterative algorithm that represents time evolution of a partial differential equation with Laplacian. We introduce the morphological Laplacian to the basic diffusion and unwrap to the deep network. The morphological filters are nonlinear operators and have parameters that are referred to as structuring elements. By the training using error back propagation, the network of the morphology can be adapted to image processing applications. Since the morphological filters is realized with addition, max and min, the error due to bit-length of data is not amplified. By this property, the morphological parts of the network are implemented in unsigned 8-bit integer with Single instruction multiple data set (SIMD) to achieve fast computation on the small micro processors. We applied the proposed network to Gaussian denoising and image completion. The results are compared with other denoising algorithm and networks.

show abstract

“…[1][2][3] So far, there have been proposed many image modeling and denoising methods. [4][5][6][7] Particularly, many algorithms in literature [4][5][6][7] have achieved good results in exploring the relationship between signal and noise, but most of them need to assume that the noise follows the uniform white Gaussian noise distribution with fixed, known intensity, and simultaneously require these noisy patches share the same or similar characteristics in some transformation domains. These methods usually achieve good denoising performance under ideal conditions.…”

mentioning

confidence: 99%

Dynamic coarse‐to‐fine ISAR image blind denoising using active joint prior learning

Xue

Feng

et al. 2021

Int J Intell Syst

View full text Add to dashboard Cite

Most existing nonblind denoising approaches assumed the noise to be homogeneous white Gaussian distribution with known intensity. However, it is difficult to know beforehand or model accurately real‐world noises with complex hybrid distribution and noise intensity. In this paper, active joint prior learning (JPL) is proposed for real‐world ISAR image blind denoising. (1) To explore strong model hierarchy and components relationship automatically, a novel graphical Dirichlet mixture process (GDMP) model is developed, where the latent representations and component hyperparameters are jointly learned from each other. (2) A multiscale joint learning strategy (MJLS) is proposed to take advantage of both the optimization‐ and discriminative learning‐based capabilities. The external noiseless, internal noisy image information and their relationships are jointly explored simultaneously. (3) Low‐rank weighted sparse learning (LWSL) is proposed to learn sparse discriminative correlation components for robust prior learning, and latent low‐rank embedding for GDMP patterns self‐adaptive inference. Extensive experimental results on ISAR image datasets demonstrate the effectiveness of the proposed model for both synthesis and real‐world noisy ISAR images, and the proposed method outperforms the state‐of‐the‐art denoising methods.

show abstract

Deep Morphological Filter Networks For Gaussian Denoising

Cited by 7 publications

References 17 publications

Quantization method for bipolar morphological neural networks

Quantization method for bipolar morphological neural networks

Deep Unrolling of Diffusion Process with Morphological Laplacian and its Implementation with SIMD Instructions

Dynamic coarse‐to‐fine ISAR image blind denoising using active joint prior learning

Contact Info

Product

Resources

About