Huge amount of work is getting done on Image colorization worldwide. This research paper proposes a model for image colorization while making use of fully automatic Convolutional Neural Network. Image colorization processes a daunting task, and this research paper proposes a relevant model for the prediction of A and B models for LAB color space and it makes a direct use the lightness channel. In this work, a pre-trained VGG-16 network was used for semantically interpreting the concepts associated with images and coloring the images. In the proposed work, the convolutional layer has been fused with the max pooling layer (higher one) of the VGG network. Architecture of the proposed model has been presented. The experimentation has been carried out with varying objective functions. LaMem experimental dataset has been used in this work in order to validate the proposed model. The proposed model is evaluated and results are visualized by histograms for true and predicted images for RGB values. Further, the proposed model has been compared with the existing models and performs better in terms of execution times (in s) for different image sizes and the results are tabulated.

中文翻译：

---

卷积神经网络实现图像着色

全球范围内有关图像着色的大量工作正在完成。该研究论文提出了一种利用全自动卷积神经网络进行图像着色的模型。图像着色是一项艰巨的任务，本文为LAB色彩空间的A和B模型的预测提出了一个相关模型，并直接使用了亮度通道。在这项工作中，使用了预训练的VGG-16网络来语义解释与图像相关的概念并为图像着色。在提出的工作中，卷积层已与VGG网络的最大池化层（较高的合并层）融合在一起。已经提出了所提出模型的体系结构。实验已经用不同的目标函数进行了。LaMem实验数据集已用于这项工作中，以验证所提出的模型。对提出的模型进行评估，并通过直方图显示真实值和预测图像的RGB值，并将结果可视化。此外，将所提出的模型与现有模型进行了比较，并且在不同图像尺寸的执行时间（以秒为单位）方面表现更好，并将结果制成表格。