End-to-End Detection of Fake Colorized Images Using CNN Framework with Multi-Colour Space Analysis

Detecting forged colorized images is crucial as advanced image editing techniques become more widespread.This study presents a Convolutional Neural Network (CNN) framework designed to detect artificially colorized images efficiently and accurately.The system integrates RGB, HSV, and YUV color spaces and incorporates color consistency metrics such as SSIM and E color difference to identify subtle manipulations.Trained on 20,000 images from ImageNet, COCO, and CelebA, the model achieves 95.7% accuracy, 94.5% precision, 96.2% recall, 95.3% F1-score, 98.2% AUC-ROC, and 93.9% MCC.The proposed model also processes images 20-30% faster than ResNet-based approaches while using less memory.This research contributes to image forensics, content verification, and security applications, offering a robust solution for detecting fake colorized images.