Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{180894,
        author = {Mohammed Sanaullah Avez and Dr. P. Vishwapathi and Syed Hamid and Mohammed Faiq Ali},
        title = {Pixvive : Reconstruction of old images using a dual-CNN pipeline for restoration and enhancement},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {12},
        number = {1},
        pages = {3047-3051},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=180894},
        abstract = {This project introduces a robust AI-based image restoration pipeline that utilizes a two-stage convolutional neural network (CNN) framework to colorize and enhance grayscale images. The pipeline’s modular architecture is designed for high efficiency and scalability, beginning with a command-line interface that ingests a grayscale input image. In the first stage, a colorization module loads a pretrained CNN model defined by a prototxt configuration and Caffe weight files. This model processes the luminance (L) channel of the input in the LAB color space, generating the corresponding chrominance (a and b) channels using learned color mappings from large-scale datasets. The output is a perceptually realistic, colorized image that preserves the structural integrity and semantic features of the original grayscale input. In the second stage, the enhancement module incorporates a pretrained Real-ESRGAN model based on the deep residual CNN architecture (RRDBNet). This model leverages residual-in-residual dense blocks to perform super-resolution upscaling, enhancing image fidelity and recovering fine details lost during earlier processing. The enhanced image is then persistently stored on disk. This system architecture underscores the sequential data flow through modular components, the distinct roles of the CNN models in transforming and refining the data, and the efficacy of combining colorization and super-resolution in a unified pipeline. Such an approach provides a scalable, end-to-end solution for restoring grayscale imagery to high-quality, colorized, and visually enriched outputs, positioning it for potential applications in archival restoration, digital content enhancement, and computational photography.},
        keywords = {Image Colourization, Deep Learning, Convolutional Neural Networks, CIE Lab Colour Space, Web API, Image Processing.},
        month = {June},
        }