4BY4(KRX,389140 KQ), the leading content AI solution provider led by CEO Junho Yun, proudly announces the publication of a research paper in the distinguished scientific journal, Scientific Reports. The paper, authored by the firm’s in-house AI research powerhouse, Pixell Lab, delves into an advanced AI model for image enhancement, setting a new standard in the realm of visual technology.
Entitled “Dual-color space network with global priors for photo retouching,” the paper explores an innovative approach to address the challenges posed by images captured in diverse lighting environments. Published in Scientific Reports, a peer-reviewed journal by Nature Portfolio, the paper exemplifies the rigor and excellence of the research, aligning with the highest standards of scientific inquiry.
Pixell Lab’s latest research tackles the complexities of images taken in environments with extreme lighting conditions, presenting a paradigm-shifting method to improve image quality. Unlike traditional AI models that often overlook variations in brightness, newly suggested model introduces an intelligent process. By first generating intermediate data of average brightness values, it fully captures color information, allowing for a more nuanced and accurate enhancement.
The key innovation lies in the implementation of two distinct networks—a “Transitional Network” to derive average brightness values and a sequential “Base Network” for image quality enhancement. This dual-color space network excels at overcoming the limitations of existing models, providing superior performance across various datasets.
Hyuncheol Kim who leads Pixell Lab at 4by4, emphasized the real-world impact of this research. “Our dual-color space network outperforms single-color space networks in diverse datasets, showcasing its effectiveness in producing visually striking and true-to-life images. The inclusion of our paper in Scientific Reports further validates the creativity and ingenuity of our image preprocessing methods.”
The proposed model is not just a theoretical breakthrough; it has already been seamlessly integrated into the firm’s AI solution, Pixell. Early market feedback attests to its capabilities, notably enhancing CCTV footage captured in low-light conditions, thereby enhancing the efficacy of criminal investigations.
Daniel Lim, Executive VP overseeing the Pixell business division, highlighted the broader implications of this research. “Beyond this featured paper, we are actively communicating our technological prowess, exemplified by our ongoing international PCT patent application for a ‘Method of improving precursor conditions.’ 4BY4 is committed to being at the forefront of innovation, setting the stage for a transformative future in visual technology.”
