Image Denoising with CNN-Based Attention DOI

Neval Karaca,

Serdar Çiftçi

Опубликована: Дек. 21, 2023

Noise removal is one of the most commonly used processes in computer vision. improves quality image, thereby improving performance vision algorithms and providing user pleasing. In this study, we aim to improve noise by adding an efficient attention module, Convolutional Block Attention Module (CBAM), Fast Flexible Denoising Network (FFDNet) model with adjustable level map as input. By CBAM module convolutions FFDNet, CNN's representational power was increased successful results were obtained. The proposed method achieved high PSNRs quantitative experiments on different datasets, qualitative observed that denoised images are close target images.

Язык: Английский

An adaptive watershed segmentation based medical image denoising using deep convolutional neural networks DOI
Ambika Annavarapu, Surekha Borra

Biomedical Signal Processing and Control, Год журнала: 2024, Номер 93, С. 106119 - 106119

Опубликована: Март 2, 2024

Язык: Английский

Процитировано

10
Enhancing Facial Recognition Accuracy in Low-Light Conditions Using Convolutional Neural Networks DOI Creative Commons

S. Swapna Rani

Deleted Journal, Год журнала: 2024, Номер 20(5s), С. 2140 - 2148

Опубликована: Апрель 13, 2024

Facial recognition technology has become increasingly everywhere in various domains, from security and surveillance to personal device authentication. However, its effectiveness can be significantly hindered low-light conditions, where images often lack sufficient illumination for accurate recognition. This study proposes a novel approach enhance facial accuracy conditions using Convolutional Neural Networks (CNNs), Deep Retinex Decomposition Network (DRDN), CenterFace algorithm. The methodology leverages CNNs robust feature extraction, while DRDN corrects by decomposing images. integrates fusion denoising layers discriminative features noise mitigation. Experimental results demonstrate remarkable improvement performance, exceeding 80% accuracy. showcases the potential of CNN-based methods with advanced techniques reliability real-world applications, particularly environments.

Язык: Английский

Процитировано

0
Image Quality Enhancement using Deep Convolutional Network DOI

Gayathri Mohan,

N Shruthi,

A R Yashaswini

и другие.

2022 International Conference on Inventive Computation Technologies (ICICT), Год журнала: 2024, Номер unknown

Опубликована: Апрель 24, 2024

People no longer tolerate images with low resolution or poor quality (color and contrast). However, certain criteria ensure the accessibility of such in both professional personal lives. Security monitoring cameras create low-resolution to optimize limited bandwidth storage space. While mobile phones are more prevalent people's daily lives than professional-grade cameras, their photography remains inferior due hardware gaps sensors chips. Lighting conditions photographic expertise also limit visual captured images. Nowadays people show an interest image enhancement improve quality. enhancing from a source color contrast challenging task. In this research, three different Convolutional Neural Networks (CNNs): Deeply-Recursive Network (DRCN), Super-Resolution (SRCNN), Residual (SRResNet) employed for enhancement. For samples were collected DIV2K dataset. Upon comparing all models, SRResNet performed well minimal Mean Squared Error (MSE) 22.05, Root (RMSE) 4.69, higher Peak Signal-to-Noise Ratio (PSNR) 37.21, along Structural Similarity Index (SSIM) 0.976.

Язык: Английский

Процитировано

0
Convolutional Autoencoder for Reconstruction of Historical Document Images: Ancient Manuscript Babad Lombok DOI Creative Commons

Fahmi Syuhada,

Asno Azzawagama Firdaus, Ana Tsalitsatun Ni’mah

и другие.

Rekayasa, Год журнала: 2024, Номер 17(1), С. 175 - 185

Опубликована: Июнь 24, 2024

The Babad Lombok is an ancient literary or manuscripts document that generally contains stories about the origins of people Lombok. This written on a lontar leaf, which in past was used to write manuscripts, letters, and documents. At present, can be seen form photos scans, so it viewed without having go museum cultural heritage site where usually exhibited. However, because this artifact has been around for hundreds years, naturally experienced fading original its scanned versions. makes text inside less clear. paper proposes automatically reconstruct/repair using neural network. type network Autoencoder Convolutional (CAE). CAE model built sequentially trained images as training data manually corrected target ground truth data. In process, two types are iteratively cropped size 64x64 along image. process results input research, each consisting 48,288 images. Testing autoencoder shows have successfully repaired, making quality clearer before reconstruction. Ultimately, proposed achieved validation accuracies 89.09% 94.57%, with corresponding loss values 0.0418 0.0226.

Язык: Английский

Процитировано

0
Deep residual learning-based denoiser for medical X-ray images DOI
Ajay Mittal,

Navdeep Kaur,

Aastha Gupta

и другие.

Evolving Systems, Год журнала: 2024, Номер unknown

Опубликована: Сен. 6, 2024

Язык: Английский

Процитировано

0
Improved Target Detection with YOLOv8 for GAN Augmented Polarimetric Images using MIRNet Denoising Model DOI Creative Commons
J. Dey,

P. Anandan,

Sonaa Rajagopal

и другие.

IEEE Access, Год журнала: 2024, Номер 12, С. 166885 - 166910

Опубликована: Янв. 1, 2024

Язык: Английский

Процитировано

0
Image Denoising with CNN-Based Attention DOI

Neval Karaca,

Serdar Çiftçi

Опубликована: Дек. 21, 2023

Noise removal is one of the most commonly used processes in computer vision. improves quality image, thereby improving performance vision algorithms and providing user pleasing. In this study, we aim to improve noise by adding an efficient attention module, Convolutional Block Attention Module (CBAM), Fast Flexible Denoising Network (FFDNet) model with adjustable level map as input. By CBAM module convolutions FFDNet, CNN's representational power was increased successful results were obtained. The proposed method achieved high PSNRs quantitative experiments on different datasets, qualitative observed that denoised images are close target images.

Язык: Английский

Процитировано

1