Modelling of chromatic contrast for retrieval of wallpaper images

Xiaohong Gao, Yuanlei Wang, Qian Yu, Alice L. L. GaoColor 2014 · DOI

TL;DR

This paper introduces CAMcc, a novel color appearance model that accounts for simultaneous chromatic contrast, a phenomenon where surrounding colors influence how we perceive a given color. This model significantly improves the accuracy of color-based image retrieval, particularly for color-rich images like wallpapers, by better capturing foreground and background color interactions.

What Problem Does This Paper Address?

Content-based image retrieval (CBIR) heavily relies on visual features like color, texture, and shape. While color is a crucial feature, existing color models often overlook how our perception of a color changes based on its surroundings – a phenomenon known as simultaneous chromatic contrast. This effect is particularly relevant in images with distinct foreground and background elements, such as wallpapers, where the background can dramatically alter the perceived hue, saturation, and lightness of foreground objects. The paper aims to bridge this gap by developing a color model that explicitly incorporates this contrast effect, leading to more perceptually accurate image retrieval.

Methods

The core contribution is the development of the CAMcc (Color Appearance Model with Colour Contrast) model. This model builds upon the CIECAM02 color appearance model but modifies its chromatic adaptation component to account for simultaneous contrast. Instead of assuming a neutral background, CAMcc treats the background color as an "adopted white" and adjusts the color perception accordingly. This is achieved by reformulating the chromatic adaptation equations to incorporate the background color's influence. For image retrieval, CAMcc is used to extract color features, which are then compared using histogram intersection to find similar images.

Datasets

The study utilizes a dataset of 654 wallpaper images collected from the Museum of Domestic Design & Architecture (MoDA) at Middlesex University. This collection spans decorative arts from the 1870s to the 1960s and is characterized by its color richness.

Key Insights and Ideas

  • Color perception is contextual: The paper strongly emphasizes that color is not perceived in isolation. The surrounding colors (background) significantly influence how we see foreground colors, a fact often neglected in standard color models.
  • CAMcc outperforms existing models: By explicitly modeling chromatic contrast, CAMcc demonstrates superior performance in retrieving visually similar images compared to established models like CIECAM02, HSI, and RGB, especially for color-intensive images.
  • Foreground focus is key: CAMcc's success stems from its ability to maintain a balance between foreground and background colors, giving more weight to foreground perception, which is often the primary focus for users when retrieving images.
  • Hue is critical for contrast: The paper highlights that the shift in hue due to simultaneous contrast is a significant factor, and CAMcc's ability to preserve hue information in retrieved images is a key advantage.

Key Results

  • Improved retrieval accuracy: CAMcc consistently outperformed CIECAM02, HSI, and RGB in retrieving images that were perceptually closer to the query images across various backgrounds.
  • Balanced foreground/background representation: CAMcc provided more balanced retrieval results by considering both foreground and background colors, unlike other models that tended to focus on dominant background tones.
  • Better hue preservation: CAMcc was particularly effective at retaining the hue content of foreground colors in the retrieved images, a crucial aspect influenced by simultaneous contrast.
  • Quantitative evaluation: While specific accuracy percentages are not explicitly stated in the abstract, the paper presents correlation coefficients (r values) in Figure 4, showing that CAMcc generally achieved better agreement with human subject estimations for lightness, colorfulness, and hue compared to CIECAM02. For instance, for red backgrounds, CAMcc achieved r=0.901 for hue, while CIECAM02 achieved r=0.682.

Limitations and Future Work

  • Effectiveness on color-intensive images: The authors note that CAMcc is most effective for "colour intensive images whereby chromatic contrast usually occurs." For images with less distinct color contrast, its performance may be similar to CIECAM02.
  • Background color determination: The current method for identifying the background color relies on histogram peak detection, which can be uncertain when multiple peaks have similar values. Future work could investigate more robust methods for background segmentation.
  • Media dependency: The study's conclusions are based on experiments conducted on an LCD monitor. Further research is needed to evaluate CAMcc's performance on other display media (e.g., reflection, CRT).
  • Integration with other features: For more comprehensive image retrieval, especially for art images where style and semantics are important, future work should explore combining color-based retrieval with other features like texture and textual descriptions.

Summary updated

Abstract

Abstract Colour remains one of the key factors in presenting an object and, consequently, has been widely applied in retrieval of images based on their visual contents. However, a colour appearance changes with the change of viewing surroundings, the phenomenon that has not been paid attention yet while performing colour‐based image retrieval. To comprehend this effect, in this article, a chromatic contrast model, CAMcc, is developed for the application of retrieval of colour intensive images, cementing the gap that most of existing colour models lack to fill by taking simultaneous colour contrast into account. Subsequently, the model is applied to the retrieval task on a collection of museum wallpapers of colour‐rich images. In comparison with current popular colour models including CIECAM02, HSI and RGB, with respect to both foreground and background colours, CAMcc appears to outperform the others with retrieved results being closer to query images. In addition, CAMcc focuses more on foreground colours, especially by maintaining the balance between both foreground and background colours, while the rest of existing models take on dominant colours that are perceived the most, usually background tones. Significantly, the contribution of the investigation lies in not only the improvement of the accuracy of colour‐based image retrieval but also the development of colour contrast model that warrants an important place in colour and computer vision theory, leading to deciphering the insight of this age‐old topic of chromatic contrast in colour science. © 2014 Wiley Periodicals, Inc. Col Res Appl, 40, 361–373, 2015
BibTeX
@inproceedings{gao2014modelling,
  author = {Xiaohong Gao and Yuanlei Wang and Qian Yu and Alice L. L. Gao},
  title = {{Modelling of chromatic contrast for retrieval of wallpaper images}},
  booktitle = {Color},
  year = {2014},
  doi = {10.1002/col.21897},
}
Figure