Masaaki Ikehara
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Masaaki Ikehara has a total of 56 co-authors. Here are the top co-authors by number of shared publications:
- Yuichi Tanaka
Graduate School of Engineering, The University of Osaka, Osaka, Japan
8 shared publications - Takuro Yamaguchi
Department of Electronics and Electrical Engineering, Keio University, Yokohama-shi, 223-8522, Japan
5 shared publications - Kei Shibasaki
Department of Electrical and Information Engineering, Faculty of Science and Technology, Keio University, Yokohama, Kanagawa, Japan
4 shared publications - Takanori Fujisawa
Department of Electronics and Electrical Engineering, Keio University, Yokohama 223-8522, Japan
4 shared publications - Ryo Abiko
Department of Electronics and Electrical Engineering, Keio University, Yokohama-shi, 223-8522, Japan
3 shared publications - Taizo Suzuki
University of Tsukuba
3 shared publications
Latest Publications
- Open Access
- 22 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 215636-215646
Consistent Fully Automatic Image Colorization Due to Bridging Fully Automatic and Point Interactive Models
Image colorization is a fundamental task in computer vision that aims to predict the missing color channels from grayscale images. In recent years, fully automatic approaches based on deep learning have become the dominant paradigm. However, these methods often produce visually unnatural results, such as color bleeding or inconsistent colorization in homogeneous regions. On the other hand, user interactive methods, such as point interactive colorization, propagate colors based on user-provided hints and tend to produce more natural and spatially consistent results. Nevertheless, when no hints are provided, the generated images may suffer from low color saturation. In this study, we propose a novel fully automatic colorization framework that combines the strengths of both paradigms: a conventional fully automatic colorization model is used as a hint generator, and a conventional point interactive colorization model is employed as a hint propagator. By treating the interactive model as a propagator within an automatic pipeline, our method ensures that the inherent colorfulness in automatic models is preserved while achieving the spatial consistency characteristic of interactive methods. Importantly, the proposed framework is fully automatic, requires no manual input, and does not necessitate retraining, as it can directly leverage existing pretrained models. We evaluated the proposed method using various fully automatic colorization models and a representative point interactive model. The results demonstrate that our method effectively reduces color inconsistencies in continuous regions and improves visual realism.
- 22 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 215636-215646
Open AccessConsistent Fully Automatic Image Colorization Due to Bridging Fully Automatic and Point Interactive Models
Image colorization is a fundamental task in computer vision that aims to predict the missing color channels from grayscale images. In recent years, fully automatic approaches based on deep learning have become the dominant paradigm. However, these methods often produce visually unnatural results, such as color bleeding or inconsistent colorization in homogeneous regions. On the other hand, user interactive methods, such as point interactive colorization, propagate colors based on user-provided hints and tend to produce more natural and spatially consistent results. Nevertheless, when no hints are provided, the generated images may suffer from low color saturation. In this study, we propose a novel fully automatic colorization framework that combines the strengths of both paradigms: a conventional fully automatic colorization model is used as a hint generator, and a conventional point interactive colorization model is employed as a hint propagator. By treating the interactive model as a propagator within an automatic pipeline, our method ensures that the inherent colorfulness in automatic models is preserved while achieving the spatial consistency characteristic of interactive methods. Importantly, the proposed framework is fully automatic, requires no manual input, and does not necessitate retraining, as it can directly leverage existing pretrained models. We evaluated the proposed method using various fully automatic colorization models and a representative point interactive model. The results demonstrate that our method effectively reduces color inconsistencies in continuous regions and improves visual realism.
- Open Access
- 9 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 158494-158503
iReWindColor v2: Cross-Window Enhanced Transformer for Point Interactive Image Colorization
Point-interactive image colorization is a task that enables users to add color to grayscale images by providing sparse color annotations (hints) at user-defined locations. These hints are then propagated across the image to generate a plausible colorization result with minimal user effort. While existing methods have demonstrated effectiveness in leveraging user input, they often suffer from excessive color diffusion and color bleeding beyond object boundaries, compromising visual fidelity. In this work, we present an extended version of our previous method, iReWindColor, which introduced a window-based Vision Transformer (ViT) framework for efficient and fine-grained colorization. The extended framework incorporates two key improvements. First, we enhance the Window Encoder by integrating Window-Connecting Convolution module, which enables information exchange across adjacent windows. This addition mitigates discontinuities at window boundaries and improves spatial coherence in the colorized output. Second, we provide a comprehensive ablation study to quantitatively assess the impact of each architectural component and design choice, offering deeper insights into the effectiveness of the proposed modifications. Our approach employs small patches to achieve detailed color propagation while maintaining computational efficiency through localized attention. The Residual Embedding mechanism remains integral to aggregating global context, whereas the proposed Encode Block, which combines Self-Attention and Convolutional Neural Networks (CNNs), strengthens inter-patch relationships and improves the utilization of user-provided hints. Experimental results on multiple benchmark datasets demonstrate that the extended model consistently outperforms the original iReWindColor as well as other state-of-the-art methods, producing visually coherent, user-aligned colorizations with improved computational efficiency.
- 9 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 158494-158503
Open AccessiReWindColor v2: Cross-Window Enhanced Transformer for Point Interactive Image Colorization
Point-interactive image colorization is a task that enables users to add color to grayscale images by providing sparse color annotations (hints) at user-defined locations. These hints are then propagated across the image to generate a plausible colorization result with minimal user effort. While existing methods have demonstrated effectiveness in leveraging user input, they often suffer from excessive color diffusion and color bleeding beyond object boundaries, compromising visual fidelity. In this work, we present an extended version of our previous method, iReWindColor, which introduced a window-based Vision Transformer (ViT) framework for efficient and fine-grained colorization. The extended framework incorporates two key improvements. First, we enhance the Window Encoder by integrating Window-Connecting Convolution module, which enables information exchange across adjacent windows. This addition mitigates discontinuities at window boundaries and improves spatial coherence in the colorized output. Second, we provide a comprehensive ablation study to quantitatively assess the impact of each architectural component and design choice, offering deeper insights into the effectiveness of the proposed modifications. Our approach employs small patches to achieve detailed color propagation while maintaining computational efficiency through localized attention. The Residual Embedding mechanism remains integral to aggregating global context, whereas the proposed Encode Block, which combines Self-Attention and Convolutional Neural Networks (CNNs), strengthens inter-patch relationships and improves the utilization of user-provided hints. Experimental results on multiple benchmark datasets demonstrate that the extended model consistently outperforms the original iReWindColor as well as other state-of-the-art methods, producing visually coherent, user-aligned colorizations with improved computational efficiency.
- 8 December 2024
- Published by Springer Nature
- p. 318-334
Image Deraining with Frequency-Enhanced State Space Model
- 8 December 2024
- Published by Springer Nature
- p. 318-334
Image Deraining with Frequency-Enhanced State Space Model
- Open Access
- 24 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 161312-161328
Faster Training of Large Kernel Convolutions on Smaller Spatial Scales
Computational requirements for training state-of-the-art neural network models are increasing on vision tasks because high computational factors have become known to be effective in improving quality. While research in the image-processing field requires a lot of trials, this trend makes proving hypotheses difficult for researchers in computationally restricted environments. Neural convolution with a wide receptive field is one of the high-computational factors with quality improvement. This study aims to accelerate the training of the large kernel convolutions by resizing both training images and convolution filters to a smaller scale. Applying this strategy requires careful training designs to replace conventional training on the target scale, and we propose four techniques to improve the quality of the trained models. In our experiment, we apply our proposals to train an image classifier model modified from RepLKNet-31B on the image classification task of the CIFAR-10, CIFAR-100, and STL-10 datasets. Our proposed framework trains almost the same models 4.62–4.91 times faster than the standard training on the target spatial scale, keeping its accuracy, and provides 2.61–2.79 times further training acceleration and stability in accuracy compared to Progressive Learning. In addition to the training acceleration, our framework can simultaneously train models for multiple scales without any scale-specific tuning, which provides scalable usage considering the computational costs.
- 24 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 161312-161328
Open AccessFaster Training of Large Kernel Convolutions on Smaller Spatial Scales
Computational requirements for training state-of-the-art neural network models are increasing on vision tasks because high computational factors have become known to be effective in improving quality. While research in the image-processing field requires a lot of trials, this trend makes proving hypotheses difficult for researchers in computationally restricted environments. Neural convolution with a wide receptive field is one of the high-computational factors with quality improvement. This study aims to accelerate the training of the large kernel convolutions by resizing both training images and convolution filters to a smaller scale. Applying this strategy requires careful training designs to replace conventional training on the target scale, and we propose four techniques to improve the quality of the trained models. In our experiment, we apply our proposals to train an image classifier model modified from RepLKNet-31B on the image classification task of the CIFAR-10, CIFAR-100, and STL-10 datasets. Our proposed framework trains almost the same models 4.62–4.91 times faster than the standard training on the target spatial scale, keeping its accuracy, and provides 2.61–2.79 times further training acceleration and stability in accuracy compared to Progressive Learning. In addition to the training acceleration, our framework can simultaneously train models for multiple scales without any scale-specific tuning, which provides scalable usage considering the computational costs.
- Open Access
- 4 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 151975-151986
Efficient and Effective Blind JPEG Image Improvement With Sequential Feature Processing
Image compression technologies such as JPEG enable efficient use of digital images. However, these technologies also degrade the quality of the image, resulting in a poor visual appearance and reduced usability. One way to solve the quality problem and achieve high efficiency in using high-quality JPEG images is to use quality improvement technologies based on artifact removal. In recent years, various methods using machine-learning-optimized neural networks have been devised to achieve high performance whereas maintaining versatility for JPEG images with various qualities. On the other hand, methods that achieve high-quality images need enormous computing costs, which is a barrier to easy use, especially when using images with large size. Given this situation, this paper proposes a method that achieves higher performance with smaller computational resources while maintaining the image quality in broader situations. The proposed method consists of a Quality Estimation Part (QE Part) that estimates the quality of the input image and Image Processing Parts (IP Parts) that process the image based on the estimated quality representation. Sequential processing is carried out by connecting multiple IP Parts in a cascade, which enables efficient and effective processing of different features. Each IP Part consists of multiple Processing Blocks, which enable effective quality improvement while maintaining efficiency. These measures enable the proposed method to achieve state-of-the-art quantitative results and better qualitative results that outperform conventional methods for images with various qualities. The code and the pre-trained models are released at https://github.com/ezumi-keio/Sequential_Processing-main.
- 4 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 151975-151986
Open AccessEfficient and Effective Blind JPEG Image Improvement With Sequential Feature Processing
Image compression technologies such as JPEG enable efficient use of digital images. However, these technologies also degrade the quality of the image, resulting in a poor visual appearance and reduced usability. One way to solve the quality problem and achieve high efficiency in using high-quality JPEG images is to use quality improvement technologies based on artifact removal. In recent years, various methods using machine-learning-optimized neural networks have been devised to achieve high performance whereas maintaining versatility for JPEG images with various qualities. On the other hand, methods that achieve high-quality images need enormous computing costs, which is a barrier to easy use, especially when using images with large size. Given this situation, this paper proposes a method that achieves higher performance with smaller computational resources while maintaining the image quality in broader situations. The proposed method consists of a Quality Estimation Part (QE Part) that estimates the quality of the input image and Image Processing Parts (IP Parts) that process the image based on the estimated quality representation. Sequential processing is carried out by connecting multiple IP Parts in a cascade, which enables efficient and effective processing of different features. Each IP Part consists of multiple Processing Blocks, which enable effective quality improvement while maintaining efficiency. These measures enable the proposed method to achieve state-of-the-art quantitative results and better qualitative results that outperform conventional methods for images with various qualities. The code and the pre-trained models are released at https://github.com/ezumi-keio/Sequential_Processing-main.
- Open Access
- 22 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 58057-58066
Multiple Adverse Weather Removal Using Masked-Based Pre-Training and Dual-Pooling Adaptive Convolution
Removing artifacts caused by multiple adverse weather, including rain, fog, and snow, is crucial for image processing in outdoor environments. Conventional high-performing methods face challenges, such as requiring pre-specification of weather types and slow processing times. In this study, we propose a novel convolutional neural network-based hierarchical encoder-decoder model that addresses these issues effectively. Our model utilizes knowledge of feature representations obtained from masked-based pre-training on a large-scale dataset. To remove diverse degradations efficiently, we employ a proposed dual-pooling adaptive convolution, which improves representational capability of weight generating network by using average pooling, max pooling, and filter-wise global response normalization. Experiments conducted on both synthetic and real image datasets show that our model achieves promising results. The performance on real images is also improved by a novel learning strategy, in which a model trained on the synthetic image dataset is fine-tuned to the real image dataset. The proposed method is notably cost-effective in terms of computational complexity and inference speed. Moreover, ablation studies show the effectiveness of various components in our method.
- 22 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 58057-58066
Open AccessMultiple Adverse Weather Removal Using Masked-Based Pre-Training and Dual-Pooling Adaptive Convolution
Removing artifacts caused by multiple adverse weather, including rain, fog, and snow, is crucial for image processing in outdoor environments. Conventional high-performing methods face challenges, such as requiring pre-specification of weather types and slow processing times. In this study, we propose a novel convolutional neural network-based hierarchical encoder-decoder model that addresses these issues effectively. Our model utilizes knowledge of feature representations obtained from masked-based pre-training on a large-scale dataset. To remove diverse degradations efficiently, we employ a proposed dual-pooling adaptive convolution, which improves representational capability of weight generating network by using average pooling, max pooling, and filter-wise global response normalization. Experiments conducted on both synthetic and real image datasets show that our model achieves promising results. The performance on real images is also improved by a novel learning strategy, in which a model trained on the synthetic image dataset is fine-tuned to the real image dataset. The proposed method is notably cost-effective in terms of computational complexity and inference speed. Moreover, ablation studies show the effectiveness of various components in our method.
- Open Access
- 25 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 146054-146064
Pose-Aware Disentangled Multiscale Transformer for Pose Guided Person Image Generation
Pose Guided Person Image Generation (PGPIG) is a task that involves generating an image of a person in a target pose, given an image in a source pose, source pose information and target pose information. Many of the existing PGPIG techniques need extra pose-related data or tasks, which cause the limitations on their applicability. In addition CNNs are used as the feature extractor which do not have long-range dependency. However, CNNs can only extract features from neighboring pixels and cannot consider image consistency. This paper introduces a PGPIG network that solves these challenges by incorporating modules that use Axial Transformers with wide receptive fields. The proposed approach disentangles the PGPIG task into two subtasks: “rough pose transformation" and “detailed texture generation." In “rough pose transformation," lower-resolution feature maps is processed by Axial Transformer-based blocks. These blocks employ an Encoder-Decoder structure, which allows the network to use the pose information well and improves the stability and performance of the training. The latter subtask employs a CNN network with Adaptive Instance Normalization. Experimental results show the competitive performance of the proposed method compared to existing methods. The proposed method achieves lowest LPIPS in Deep Fashion dataset and FID in Market-1501 dataset. Remarkably, despite the great results obtained, the number of parameters of the proposed network is significantly less in contrast to existing methods.
- 25 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 146054-146064
Open AccessPose-Aware Disentangled Multiscale Transformer for Pose Guided Person Image Generation
Pose Guided Person Image Generation (PGPIG) is a task that involves generating an image of a person in a target pose, given an image in a source pose, source pose information and target pose information. Many of the existing PGPIG techniques need extra pose-related data or tasks, which cause the limitations on their applicability. In addition CNNs are used as the feature extractor which do not have long-range dependency. However, CNNs can only extract features from neighboring pixels and cannot consider image consistency. This paper introduces a PGPIG network that solves these challenges by incorporating modules that use Axial Transformers with wide receptive fields. The proposed approach disentangles the PGPIG task into two subtasks: “rough pose transformation" and “detailed texture generation." In “rough pose transformation," lower-resolution feature maps is processed by Axial Transformer-based blocks. These blocks employ an Encoder-Decoder structure, which allows the network to use the pose information well and improves the stability and performance of the training. The latter subtask employs a CNN network with Adaptive Instance Normalization. Experimental results show the competitive performance of the proposed method compared to existing methods. The proposed method achieves lowest LPIPS in Deep Fashion dataset and FID in Market-1501 dataset. Remarkably, despite the great results obtained, the number of parameters of the proposed network is significantly less in contrast to existing methods.
- Open Access
- 1 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 137495-137505
Enhanced Unpaired Image-to-Image Translation via Transformation in Saliency Domain
Unpaired image to image translation is the task of converting images in unpaired datasets. The primary goal of the task is to translate a source image into the image aligned with the target domain while keeping the fundamental content. Existing researches have introduced effective techniques to translate images with unpaired datasets, focusing on preserving the fundamental content. However, these techniques have limitations in dealing with significant shape changes and preserving backgrounds that should not be transformed. The proposed method attempts to address these problems by utilizing the saliency domain for translation and simultaneously learning the translation in the saliency domain as well as in the image domain. The saliency domain represents the shape and position of the main object. The explicit learning of transformations within the saliency domain improves network’s ability to transform shapes while maintaining the background. Experimental results show that the proposed method successfully addresses the problems of unpaired image to image translation and achieves competitive metrics with existing methods.
- 1 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 137495-137505
Open AccessEnhanced Unpaired Image-to-Image Translation via Transformation in Saliency Domain
Unpaired image to image translation is the task of converting images in unpaired datasets. The primary goal of the task is to translate a source image into the image aligned with the target domain while keeping the fundamental content. Existing researches have introduced effective techniques to translate images with unpaired datasets, focusing on preserving the fundamental content. However, these techniques have limitations in dealing with significant shape changes and preserving backgrounds that should not be transformed. The proposed method attempts to address these problems by utilizing the saliency domain for translation and simultaneously learning the translation in the saliency domain as well as in the image domain. The saliency domain represents the shape and position of the main object. The explicit learning of transformations within the saliency domain improves network’s ability to transform shapes while maintaining the background. Experimental results show that the proposed method successfully addresses the problems of unpaired image to image translation and achieves competitive metrics with existing methods.
- Open Access
- 4 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 117520-117527
UNet Based Multi-Scale Recurrent Network for Lightweight Video Deblurring
With the recent widespread use of smartphones and digital video cameras, the opportunities to handle digital video have increased significantly. However, despite improvements in the performance of the hardware, the captured video often contains information that is not necessary for the purpose of the video. In particular, factors such as camera shake and object movement can cause blurring in video. So, we propose a method to deblur a video by software processing of the video after shooting. In conventional methods, video deblurring is often performed using a network whose main task is video super-resolution. However, in super-resolution, the size of the input and output images are different, while the input and output images are the same size in deblurring. In the case of deblurring, the input images are input to the network after a simple downsampling process, which is not optimized for the same size as the input and output images. Therefore, the proposed method constructs a multi-scale network based on UNet. The UNet-based network is a successful method for single image deblurring. Because a video is a sequence of multiple images, we use a method that has been successful in single image deblurring. Furthermore, we add improvements to the network based on the structure of MPRNet. The feature extraction modules of the bottom and the second stage of the network are replaced with a single-stage UNet. These improvements resulted in a 34.80dB in PSNR and 0.973 in SSIM on the GoPro dataset despite about 75% of the FLOPs of BasicVSR++ and 3% of the FLOPs of VRT. On the DVD dataset, the proposed model achieved 34.36dB in PSNR and 0.966 in SSIM. Further ablation studies show the effectiveness of various components in our proposed model.
- 4 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 117520-117527
Open AccessUNet Based Multi-Scale Recurrent Network for Lightweight Video Deblurring
With the recent widespread use of smartphones and digital video cameras, the opportunities to handle digital video have increased significantly. However, despite improvements in the performance of the hardware, the captured video often contains information that is not necessary for the purpose of the video. In particular, factors such as camera shake and object movement can cause blurring in video. So, we propose a method to deblur a video by software processing of the video after shooting. In conventional methods, video deblurring is often performed using a network whose main task is video super-resolution. However, in super-resolution, the size of the input and output images are different, while the input and output images are the same size in deblurring. In the case of deblurring, the input images are input to the network after a simple downsampling process, which is not optimized for the same size as the input and output images. Therefore, the proposed method constructs a multi-scale network based on UNet. The UNet-based network is a successful method for single image deblurring. Because a video is a sequence of multiple images, we use a method that has been successful in single image deblurring. Furthermore, we add improvements to the network based on the structure of MPRNet. The feature extraction modules of the bottom and the second stage of the network are replaced with a single-stage UNet. These improvements resulted in a 34.80dB in PSNR and 0.973 in SSIM on the GoPro dataset despite about 75% of the FLOPs of BasicVSR++ and 3% of the FLOPs of VRT. On the DVD dataset, the proposed model achieved 34.36dB in PSNR and 0.966 in SSIM. Further ablation studies show the effectiveness of various components in our proposed model.
- Open Access
- 31 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 81466-81476
Toward High-Quality Real-Time Video Denoising With Pseudo Temporal Fusion Network
With the increasing availability of high-resolution video recording and streaming, there is a need for fast and high-quality video denoising methods that can handle high-resolution videos. However, many existing methods fail to achieve high-quality denoising performance and lightweight at the same time. This paper proposes a video denoising network, Pseudo Temporal Fusion Network (PTFN), that satisfies these requirements. PTFN adopts a new Pseudo Temporal Fusion (PTF) module that captures pseudo-temporal relationships between video frames in combination with the Temporal Shift Module. PTFN also adopts a modern ConvBlock paradigm that breaks away from the classical ConvBlock paradigm, contributing to denoising performance and lightweight. PTFN achieves better performance than existing video denoising methods in terms of both video quality and lightweight. Specifically, PTFN has only about 16.7% of the computational cost of existing lightweight state-of-the-art methods, while it improves denoising performance. PTFN is also superior in terms of memory consumption. It can process 1080p videos with a GPU with 24 GB RAM. In addition, a lighter version (PTFN Half) can process 2K videos at high speed under the same conditions.
- 31 July 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 81466-81476
Open AccessToward High-Quality Real-Time Video Denoising With Pseudo Temporal Fusion Network
With the increasing availability of high-resolution video recording and streaming, there is a need for fast and high-quality video denoising methods that can handle high-resolution videos. However, many existing methods fail to achieve high-quality denoising performance and lightweight at the same time. This paper proposes a video denoising network, Pseudo Temporal Fusion Network (PTFN), that satisfies these requirements. PTFN adopts a new Pseudo Temporal Fusion (PTF) module that captures pseudo-temporal relationships between video frames in combination with the Temporal Shift Module. PTFN also adopts a modern ConvBlock paradigm that breaks away from the classical ConvBlock paradigm, contributing to denoising performance and lightweight. PTFN achieves better performance than existing video denoising methods in terms of both video quality and lightweight. Specifically, PTFN has only about 16.7% of the computational cost of existing lightweight state-of-the-art methods, while it improves denoising performance. PTFN is also superior in terms of memory consumption. It can process 1080p videos with a GPU with 24 GB RAM. In addition, a lighter version (PTFN Half) can process 2K videos at high speed under the same conditions.