Black & White Images Colorization Using Faster R-CNN

EasyChair Preprint 9802

Abstract

Black and white photos are an essential part of our cultural legacy, documenting moments in time that might otherwise be lost to history. Unfortunately, the lack of color might make it difficult to enjoy these photographs, and colorization methods can be costly and time-consuming. In this research, we offer a unique method for colorizing black-and-white photos using the Faster R-CNN algorithm. Our method entails first training a deep neural network to recognize and localize objects in a black-and-white image, and then using these item positions to build a colorized version of the image. Using the Faster R-CNN technique, we recognize objects inside the black and white image and use the color information from the found items to colorize the rest of the image. Using a collection of black and white photographs, we show the efficiency of our technique and obtain maximum colorization accuracy. Our method colorize black and white photographs quickly and efficiently, making them more accessible and interesting to a larger audience. We assess our method's performance using a variety of metrics, including peak signal-to-noise ratio (PSNR), and demonstrate that it creates high-quality colorized pictures with a high degree of precision.

Keyphrases: Black & White, Deep Neural Networks, Faster RCNN, Image Colorization, PSNR, pre-trained

@booklet{EasyChair:9802,
  author    = {S Kingsley and V C Sai Santhosh and P Mohamed Rasik and J Ragav Krishna and S M Sameer Sheriff},
  title     = {Black & White Images Colorization Using Faster R-CNN},
  howpublished = {EasyChair Preprint 9802},
  year      = {EasyChair, 2023}}