Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
156 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
45 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Comicolorization: Semi-Automatic Manga Colorization (1706.06759v4)

Published 21 Jun 2017 in cs.CV and cs.GR

Abstract: We developed "Comicolorization", a semi-automatic colorization system for manga images. Given a monochrome manga and reference images as inputs, our system generates a plausible color version of the manga. This is the first work to address the colorization of an entire manga title (a set of manga pages). Our method colorizes a whole page (not a single panel) semi-automatically, with the same color for the same character across multiple panels. To colorize the target character by the color from the reference image, we extract a color feature from the reference and feed it to the colorization network to help the colorization. Our approach employs adversarial loss to encourage the effect of the color features. Optionally, our tool allows users to revise the colorization result interactively. By feeding the color features to our deep colorization network, we accomplish colorization of the entire manga using the desired colors for each panel.

Citations (104)

Summary

  • The paper introduces Comicolorization, a semi-automatic tool that leverages CNNs with adversarial loss to achieve consistent, vivid manga colorization.
  • The methodology processes entire manga pages by extracting color features from reference images and allowing interactive user adjustments for precise output.
  • The paper shows that reducing manual colorization enhances publishing efficiency and broadens manga accessibility to international audiences.

Comicolorization: Semi-Automatic Manga Colorization

The paper introduces Comicolorization, a semi-automatic tool designed to enhance the traditional Japanese art form of manga by providing a method for coloring monochrome panels. This research stands out for its approach in tackling the intricacies of manga colorization, a process that involves the delicate balance of maintaining character consistency across numerous pages while introducing vibrant hues typically absent in classical manga.

Methodology and Innovations

The Comicolorization tool distinguishes itself by focusing on the colorization of entire manga pages rather than isolated panels, addressing the inherent color ambiguity problem. This challenge arises due to the visual similarity between different characters or the diverse appearances of the same characters across various panels. The methodology employs convolutional neural networks (CNNs) augmented with adversarial loss to produce more vivid and consistent character colors across pages.

The process begins by receiving monochrome manga pages and reference images, from which color features are extracted. These features guide the coloring process, aided by a deep colorization network that ensures consistent color application for similar characters. A notable aspect of this system is its semi-automatic nature, where users can interactively adjust the colorization output, allowing for manual fine-tuning via color dots and histogram modifications.

Key Results

The paper presents impressive results that highlight the system's capacity to apply vibrant colors consistently across manga panels. The authors compare their approach to existing fully automatic methods, noting that Comicolorization mitigates typical issues such as dull color palettes or inconsistent character shading.

From a performance standpoint, the system significantly reduces the manual effort required for manga artists, offering a practical balance between full automation and manual colorization. By leveraging adversarial loss, the tool produces colorized images with enhanced vividness, further distinguishing it from standard approaches.

Implications and Future Work

Practically, Comicolorization opens doors for manga publishing by reducing the time and costs associated with manual colorization. This could enhance international appeal and accessibility, as colorized manga may attract broader audiences unused to monochrome illustrations.

Theoretically, this paper contributes to the growing body of work on image colorization via deep learning, especially in domains where color consistency is critical. The blend of CNNs with adversarial networks proposes a framework that could be adapted for other forms of graphic art requiring precise color consistency.

Future developments could explore the integration of more complex user inputs or expand the dataset used for training to include a wider variety of art styles and character designs. Additionally, expanding the scalability of the system to handle more diverse manga styles with minimal user intervention might be an avenue worth exploration.

In summary, Comicolorization makes a substantive contribution to the field of image processing and presents a valuable tool for manga artists and publishers. Its combination of technological sophistication and practical application offers significant potential for both current use cases and future adaptations in related industries.