Author(s): Sachin Choudhari, Bhushan Rajani, Aditya Jethani, Kashish Dhongde, Kshitij Wanjari, Saraswati Dudani and Shrawani Petle

Abstract: Image coloring is an important task in computer vision, which includes color restoration of gray images. The traditional approach must be manually intervened, so the process time is consumed and dependent on expert knowledge. Nevertheless, the achievements of the deep learning field are possible to automatically paint, greatly reducing human efforts and increasing accuracy. This project presents a web application based on a plisk that performs automatic images using a pre -trained OpenCV in -depth education model. This model works in the color space of the laboratory where L channel (gray input) is processed, and the missing color channel A and B (color components) are predicted using deep training of the family unexplored neural network (CNN). This system implements a web platform that allows users to load gray images, and then processed through coloring models to generate realistic color output. The application provides an unbreakable user experience so that the user can see and load the processed image. Productivity assessment using structural similarity index (SSIM) and peak signal/ noise attitude (PSNR) is visually consistent and confirmed the effect of the model in the creation of the exact color for the situation. This system has actual applications in various fields, including historical restoration of photography, digital content improvement, medical visualization and media processing. The result shows that coloring based on deep learning can produce high quality results at a minimum cost of calculation. In the future, the improvement includes the user model training of deep learning, the output speed optimization for real -time processing, and the GENES (General Networks) to improve the quality of the color image and the further improvement of realism.

DOI: 10.33545/27076571.2025.v6.i1b.137

Pages: 95-98 | Views: 454 | Downloads: 112

Download Full Article: Click Here