An extended variational image decomposition model for color image enhancement

Xixi Jia, Xiangchu Feng, Weiwei Wang, Lei Zhang

Research output: Journal article publicationJournal articleAcademic researchpeer-review

4 Citations (Scopus)

Abstract

Variational image decomposition model has been widely used in computer vision and computational photography to decompose an image into a luminance layer and a reflectance layer. Existing method perform decomposition either on the intensity channel of a color image or on each of the RGB channels separately. However, these methods fail to exploit effectively the correlation among color channels, and will lose much color edges and textures during the decomposition process. This paper presents an extended variational image decomposition model, which works directly on the color image and decomposes it into a luminance layer, a reflectance layer and a color layer by solving a total variation minimization problem. The luminance layer is piecewise smooth containing the general object appearance, the reflectance layer is piecewise constant representing some structure information. The color layer accounts for distinct color structure in each color channel and reveals the color variation of the image. By manipulating the three layers, the proposed decomposition model can be applied to a variety of color image decomposition applications such as high dynamic range tone mapping and contrast enhancement. Our experimental results validate the highly competitive performance of the proposed method in comparison with state-of-the-arts.

Original languageEnglish
Pages (from-to)216-228
Number of pages13
JournalNeurocomputing
Volume322
DOIs
Publication statusPublished - 17 Dec 2018

Keywords

  • HDR tone mapping
  • Image contrast enhancement
  • Image decomposition
  • Total variation

ASJC Scopus subject areas

  • Computer Science Applications
  • Cognitive Neuroscience
  • Artificial Intelligence

Cite this