Automatic colorization of Black-White images using
deep learning is a technique to colorize black-white images without
involvement of a human. Conventional techniques used for
colorizing images need human intervention, which is timeconsuming.
The project deals with deep learning techniques to
automatically colorize black scale images. The proposed technique
uses deep convolutional neural networks and has a number of
advantages. The technique will reduce manual work, speed up
the process and improve the accuracy. Automatic colorization
techniques using ConvNet finds applications in various domains
such as astronomy, electron microscopy, and archaeology.
Conventional approach to achieve colorization included
regression-based model, graph cut algorithm etc. Proposed model
is a classification based technique but uses regression model as
the base line model. Designed system consists of training and
testing phases. Feature extraction and pixel-mapping from the
input coloured image results in training of the system. In the
testing phase the system is provided with black scale input images
to check the accuracy of colorization of these images. This
technique can be used to eliminate the need of expensive image
transferring equipments for astronomical images and to speed up
the process of conversion of legacy images to modern coloured
images, thus reducing manual effort needed by utilizing deep
learning techniques.
Keywords—Deep Learning,Automatic Colorization, Black Scale
Images.