PIXELL AI:
Deep Learning based
Image Quality Enhancement and upscaling
Our deep learning AI model, trained on tens of thousands of ultra-high-definition videos, demonstrates the most advanced level of image enhancement performance.
Model Training
& Advancement
4BY4's team of image quality experts who has been collaborating with leading global display companies to set and define the standards for 'optimal image quality' team up with top-class AI researchers. Together, they train and refine models to produce the best results. The quality data provision and continuous feedback from image quality experts on output videos, along with various research and experiments by the internal AI researchers, enables a faster and more precise enhancement of solutions.
Core Technology
Core Technology
PIXELL AI
Deep Learning based Image Quality Enhancement and upscaling
The technology optimizes each element that constitutes image quality, such as noise, sharpness, and color, to achieve the closest possible quality to human vision.
Our deep learning AI models, trained on tens of thousands of ultra-high-definition videos, demonstrates the most advanced level of image enhancement performance.
Dominant
Performance
Compared to existing image enhancement AI models, PIXELL AI demonstrates overwhelming performance differences across all areas that determine image quality: saturation, contrast, noise, and sharpness.
[출처] Chen et. al “Deep Retinex Decomposition for Low-Light Enhancement”, British Machine Vision Conference (BMVC) 2018 | Guo et. al “Zero-Reference Deep Curve Estimation for Low-Light Image Enhancement”, IEEE Conference on Computer Vision Pattern Recognition (CVPR) 2020 | Kim et. al “Accurate Image Super-Resolution Using Very Deep Convolutional Networks”, IEEE Conference on Computer Vision Pattern Recognition (CVPR) 2016 | Cho et. al “Effective Five Directional Partial Derivatives-Based Image Smoothing and a Parallel Structure Design”, IEEE Trans. on Image Processing (TIP) 2016 | Shin et. al “Region-Based Dehazing via Dual-Supervised Triple-Convolutional Network”, IEEE Trans. on Multimedia (TMM) 2020 | Zhang et. al “Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising”, IEEE Trans. on Image Processing (TIP) 2017 | Zhang et. al “Learning Deep CNN Denoiser Prior for Image Restoration”, IEEE Conference on Computer Vision Pattern Recognition (CVPR) 2017 | Zhang et. Al “FFDNet: Toward a Fast and Flexible Solution for CNN-based Image Denoising”, IEEE Trans. on Image Processing (TIP) 2018
Depending on the features of the original source video and the degree of quality degradation, a comprehensive process is undertaken, including upscaling, noise reduction, color enhancement, sharpness improvement, and brightness adjustment, to achieve results that closely resemble what is seen with the naked eye.
Scientific Recognition
Through collaborative research with PIXELL Lab, the Graduate School of Advanced Imaging Science at Chung-Ang University (AI Graduate School), and the AI startup AI Nation, we continuously enhance our core technologies in the field of image quality AI. We are strengthening our tech leadership by submitting papers to prestigious academic journals and filing patents.
Publication
Patent Registration
Model Training & Advancement
4BY4's team of image quality experts who has been collaborating with leading global display companies to set and define the standards for 'optimal image quality' team up with top-class AI researchers. Together, they train and refine models to produce the best results.
The quality data provision and continuous feedback from image quality experts on output videos, along with various research and experiments by the in-house AI researchers, enables a faster and more precise enhancement of solutions.
Dominant Performance
Compared to existing image enhancement AI models, PIXELL AI demonstrates overwhelming performance differences across all areas that determine image quality: saturation, contrast, noise, and sharpness.
[출처] Chen et. al “Deep Retinex Decomposition for Low-Light Enhancement”, British Machine Vision Conference (BMVC) 2018 | Guo et. al “Zero-Reference Deep Curve Estimation for Low-Light Image Enhancement”, IEEE Conference on Computer Vision Pattern Recognition (CVPR) 2020 | Kim et. al “Accurate Image Super-Resolution Using Very Deep Convolutional Networks”, IEEE Conference on Computer Vision Pattern Recognition (CVPR) 2016 | Cho et. al “Effective Five Directional Partial Derivatives-Based Image Smoothing and a Parallel Structure Design”, IEEE Trans. on Image Processing (TIP) 2016 | Shin et. al “Region-Based Dehazing via Dual-Supervised Triple-Convolutional Network”, IEEE Trans. on Multimedia (TMM) 2020 | Zhang et. al “Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising”, IEEE Trans. on Image Processing (TIP) 2017 | Zhang et. al “Learning Deep CNN Denoiser Prior for Image Restoration”, IEEE Conference on Computer Vision Pattern Recognition (CVPR) 2017 | Zhang et. Al “FFDNet: Toward a Fast and Flexible Solution for CNN-based Image Denoising”, IEEE Trans. on Image Processing (TIP) 2018
Depending on the features of the original source video and the degree of quality degradation,
a comprehensive process is undertaken, including upscaling, noise reduction, color enhancement, sharpness improvement, and brightness adjustment, to achieve results that closely resemble what is seen with the naked eye.
Scientific Recognition
Through collaborative research with PIXELL Lab, the Graduate School of Advanced Imaging Science at Chung-Ang University (AI Graduate School), and the AI startup AI Nation,
we continuously enhance our core technologies in the field of image quality AI.
We are strengthening our tech leadership by submitting papers to prestigious academic journals and filing patents.
Publication
Patent Registration
| Name | Registration Number | Remarks | |
Patent Registration | Method and device for Deep learning-based video noise reduction | 10-2670870 | |
| Method for Edge Enhancement Using Depth Gradient Priors Based on L0 Smoothing | 10-2342940 | PCT International Application in Progress | |
| Method for Image Quality Correction Using Input Image Preprocessing | 10-2245835 | ||
| Method for Image Output Based on Artificial Intelligence with Improved Quality | 10-2054453 | ||
| Method for Improving the Quality of Ultra-High Resolution Video Content | 10-2058355 | ||
| Method for Digital Video Noise Reduction | 10-2194447 | ||
| Method for Enhancing Sharpness in Digital Images | 10-2164998 | ||
| Image Segmentation Device and Method Based on Continuous Frames of a Still Scene | 10-1223046 | ||
| Color Image Correction Method Using an Improved Image Generation Model and Adaptive Filter | 10-1242069 | ||
| Color Image Correction Method Using a Modified Image Generation Model | 10-1242070 | ||
| SIM-Embedded Storage Medium | 10-2123312 | ||
| Locking Mechanism Between Televisions for Demonstration and Promotion and USB Memory for Video Provision | 10-1981033 |
Ⓒ2024 4BY4 Inc,.
12-14F, 479, Gangnam-daero, Seocho-gu, Seoul, Korea 06541
문의: 4by4inc@4by4inc.com / IR: ir@4by4inc.com
Family site
Ⓒ2024 4BY4 Inc.
12-14F, 479, Gangnam-daero, Seocho-gu, Seoul, Korea 06541
문의: 4by4inc@4by4inc.com / IR: ir@4by4inc.com
Family site
