Abstract :

Image colorization is an important problem in computer vision that focuses on converting grayscale images into realistic color images using learning-based techniques. Traditional approaches depend on manual intervention or handcrafted features and often fail to generalize for complex scenes. This work presents an automatic image colorization system using a Deep Convolutional Neural Network (DCNN), which learns meaningful color representations directly from large image datasets. The model extracts high-level contextual information, semantic regions, and object relationships to generate visually consistent color outputs. The proposed method reduces human effort, improves accuracy, and enhances realism in reconstructed images. Experimental results demonstrate better visual quality and robustness when compared with classical colorization methods. This approach is suitable for applications such as photo restoration, digital forensics, entertainment, and medical imaging