Models and algorithms for determining dominant colors in a digital image

Identifying dominant colors in an image is crucial for various applications, including color theme generation, image categorization, content-based search, and graphic information analysis. In this study, we analyze existing methods and models used to determine dominant colors, highlighting their advantages and limitations. In the field of graphic design, clustering models such as K-means and dE-means prove to be effective tools for extracting dominant colors, especially when the primary goal is to identify visually prominent hues. For medical or industrial image analysis, however, percep­tual models or algorithms based on neural networks offer superior performance. These models are better equipped to account for human color perception, ensuring that color extraction aligns with real-world observations and can be applied more effectively in fields where precision is critical. Modern algorithms like Convolutional Neural Networks (CNNs) can assess the quality of color themes and dynamically generate color palettes in various sizes. This study justifies the application of the ICaS color space, particularly for processing and extracting dominant colors with higher accuracy across different contexts. An orthogonal model for identifying dominant colors is proposed, which incorporates the chromatic characteristics of colors, making it more adaptable across a wide range of design and analytical applications. As part of the model’s development, boundary values for hue were established to segment colors effectively, ensuring efficient and accurate color extraction. The findings are expected to contribute to further advancements in computer vision and digital design by providing a basis for developing automated systems capable of analyzing images and supporting decision-making processes in printing.

Keywords: model, dominant colors, color scheme, image, color space, chromaticity, color tone.

doi: 10.32403/0554-4866-2025-1-89-94-102