Next Article in Journal
In Situ Ceramic Phase Reinforcement via Short-Pulsed Laser Cladding for Enhanced Tribo-Mechanical Behavior of Metal Matrix Composite FeNiCr-B4C (5 and 7 wt.%) Coatings
Next Article in Special Issue
A New Method for Camera Auto White Balance for Portrait
Previous Article in Journal
Using AI to Reconstruct and Preserve 3D Temple Art with Old Images
Previous Article in Special Issue
Human-Machine Interaction: A Vision-Based Approach for Controlling a Robotic Hand Through Human Hand Movements
 
 
Article
Peer-Review Record

New Approach to Dominant and Prominent Color Extraction in Images with a Wide Range of Hues

Technologies 2025, 13(6), 230; https://doi.org/10.3390/technologies13060230
by Yurii Kynash * and Mariia Semeniv
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Technologies 2025, 13(6), 230; https://doi.org/10.3390/technologies13060230
Submission received: 28 April 2025 / Revised: 28 May 2025 / Accepted: 1 June 2025 / Published: 4 June 2025
(This article belongs to the Special Issue Image Analysis and Processing)

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1. The paper should include a quantitative comparison between ICaS and traditional color spaces (e.g., LAB, HSV) in terms of gamut coverage and clustering accuracy, supplemented by visual examples to demonstrate ICaS’s superiority in handling wide-range hues and saliency detection.
2.  Conduct a user study where participants rate the aesthetic quality of palettes generated by OSM versus other methods, assessing harmony, appeal, and relevance to provide human-centric validation beyond technical metrics like ΔE₀₀.
3.  Analyze and compare the runtime of OSM with baseline methods (KM, FCM) to assess scalability, and suggest optimizations (e.g., GPU acceleration) for real-time applications or large datasets.
4.  Evaluate OSM’s robustness on challenging images (e.g., underwater, low-light) to demonstrate its adaptability and identify limitations in handling muted colors or poor lighting conditions.
5.  Explain how the saliency threshold (e.g., 24) impacts results, propose adaptive thresholding for varied content, and visualize saliency maps with intermediate palettes to improve reproducibility and pipeline transparency.

Author Response

We are sincerely grateful to the reviewer for his time and comments.

1. The paper should include a quantitative comparison between ICaS and traditional color spaces (e.g., LAB, HSV) in terms of gamut coverage and clustering accuracy, supplemented by visual examples to demonstrate ICaS’s superiority in handling wide-range hues and saliency detection.

Response. Thank you for your good point about the need to quantitatively compare the coverage of the ICaS color space with traditional spaces (Lab, HSV) and their ability to cluster. To compare the color coverage of the ICaS and CIE Lab spaces, we constructed two-dimensional color grids with a fixed achromatic component. Colors on the planes have the same brightness (I or L) but different chromatic coordinates. The obtained results visually demonstrate the wider color coverage of ICaS. The results are presented in the article.

2. Conduct a user study where participants rate the aesthetic quality of palettes generated by OSM versus other methods, assessing harmony, appeal, and relevance to provide human-centric validation beyond technical metrics like ΔE₀₀.

Response. Thank you very much for your question and for your good recommendation regarding the need for human-centered validation of color clustering methods that goes beyond technical metrics such as ΔE₀₀.
To fulfill this task, we conducted a visual experiment with the participation of users who evaluated the aesthetic quality of color palettes generated by OSM and other compared algorithms (K-means, FCM, KML, Shift Mean). The results are presented in the article. Histograms of the distribution of expert opinions by the criteria of harmony, attractiveness, and relevance for each of the five images are constructed.

3. Analyze and compare the runtime of OSM with baseline methods (KM, FCM) to assess scalability, and suggest optimizations (e.g., GPU acceleration) for real-time applications or large datasets.

Response. Thank you very much for your good point about the need to evaluate the performance of the OSM method in comparison to the basic approaches (K-means, FCM), as well as for the suggestion of potential optimization for scalable applications.
For the purpose of performance analysis, we added the measurement of the execution time of clustering algorithms for the task of determining the dominant colors. The testing was performed in the Google Colab environment using the start_time = time.time() command and the corresponding end_time fixation.  The results are presented in the article.

4. Evaluate OSM’s robustness on challenging images (e.g., underwater, low-light) to demonstrate its adaptability and identify limitations in handling muted colors or poor lighting conditions.

Response. Thank you for your question about the robustness of the OSM algorithm to images with challenging conditions, such as underwater scenes or low light.
This study focused on analyzing images with an extended color range, which provides the right conditions for highlighting dominant and prominent colors. Therefore, images with muted or deformed colors (e.g., due to insufficient lighting) were not included in the main experimental set.
At the same time, it should be noted that the OSM algorithm provides customizable parameters that can be adapted to such challenges:
- The value of the achromatic component allows you to filter or preserve almost gray (muted) colors.
- The chromaticity (saturation) thresholds can be lowered to accommodate low-intensity colors, which is especially important for low-light scenes.
These possibilities open up the prospect of adapting the OSM method to images with less contrast or complex lighting conditions. This issue is the subject of our further work, in particular in the context of using OSM in extended environments, including processing photos in a natural or technical context (for example, underwater photography or night photography).

5. Explain how the saliency threshold (e.g., 24) impacts results, propose adaptive thresholding for varied content, and visualize saliency maps with intermediate palettes to improve reproducibility and pipeline transparency.

Response. Thank you for your attention to the important aspect of choosing a salience threshold and its impact on the results.
In this study, the value of the threshold (24) was determined empirically based on the target characteristics of the palettes and visual analysis on different images. To demonstrate the effect of this parameter, we provide examples of images with different salience thresholds, which allows us to visually observe the change in the number and type of colors selected in the palettes.
We agree with your proposal to adapt the threshold based on the characteristics of the content. This is a promising area for further research.

Reviewer 2 Report

Comments and Suggestions for Authors
  1. Elaborate on the advantages of ICaS over standard spaces like Lab or HSV, with special attention beyond orthogonality. Include some comparison results or a short theoretical justification on the reasons for color separation, perceptual uniformity, or computational efficiency.
  2. Provide reasons for choosing the six sectors. Would adaptable techniques (e.g., silhouette score, elbow method) increase the image type versatility?

  3. Cite other approaches besides Spectral Residual, such as Deep Gaze, MLNet, or U-Net, and describe how you plan to test and improve salient color selection.

  4. Augment the dataset or implement cross-validation utilizing large publicly available datasets, such as the MIT1003 or AVA datasets, in order to test generalization.

  5. Think about adding user evaluation for perceptual quality scoring aesthetics and ranking them subjectively, or compute diversity scores based on entropy and distinctiveness perceptually.

  6. Contrast with unsupervised approaches such as Mean Shift, DBSCAN, or Self-Organizing Maps (SOM).

  7. Assessing temporal performance, or computational complexity of the OSM method.

     

 

Author Response

We thank the reviewer for his comments that helped us improve the article.

1. Elaborate on the advantages of ICaS over standard spaces like Lab or HSV, with special attention beyond orthogonality. Include some comparison results or a short theoretical justification on the reasons for color separation, perceptual uniformity, or computational efficiency.

Response. Thank you for your good point about the feasibility of using ICaS. We chose the ICaS color space because of a number of advantages that go beyond component orthogonality. ICaS allows us to clearly separate intensity (I) from color information, which improves the manageability of selecting dominant colors for salience. The color coverage of the ICaS and CIE Lab spaces is compared. The results are presented in the article.

2. Provide reasons for choosing the six sectors. Would adaptable techniques (e.g., silhouette score, elbow method) increase the image type versatility?

Response. The choice of six sectors is determined by the number of primary colors of additive (RGB) and subtractive (CMY) synthesis that provide maximum coverage of the color space in digital images. In the ICaS space, the vectors of these six colors form a hexagon that reflects the boundaries of color coverage. This structure allows you to effectively segment the color space.
We agree that adaptive methods, such as silhouette estimation or the elbow method, can increase the versatility of the algorithm when working with different types of images. In future research, we plan to implement a dynamic selection of the number of sectors depending on the color complexity of the image, which will allow us to better adapt the method to the specifics of the image.

3. Cite other approaches besides Spectral Residual, such as Deep Gaze, MLNet, or U-Net, and describe how you plan to test and improve salient color selection.

Response. At this stage of the study, we focused on using the spectral residual method, which is easy to implement and sufficient for initial analysis of salient colors. In our opinion, it would be interesting to conduct future research using proposed models such as Deep Gaze, MLNet or U-Net, which should improve the efficiency of the OSM method.

4. Augment the dataset or implement cross-validation utilizing large publicly available datasets, such as the MIT 1003 or AVA datasets, in order to test generalization.

Response. Thank you for the recommendation to expand the dataset. We chose the most colorful images from the MIT set and applied the methods for determining dominant and prominent colors used in this study. [http://saliency.mit.edu/downloads.html]

5. Think about adding user evaluation for perceptual quality scoring aesthetics and ranking them subjectively, or compute diversity scores based on entropy and distinctiveness perceptually.

Response. Thank you very much for your comment.
We have conducted a visual experiment with users who evaluated the aesthetic quality of color palettes generated by OSM and other comparable algorithms (K-means, FCM, KML, Shift Mean). The results of the experiment are included in the article.

6. Contrast with unsupervised approaches such as Mean Shift, DBSCAN, or Self-Organizing Maps (SOM).

Response. A comparison with one of the unsupervised approaches, Mean Shift, is made. The results are presented in the article.

7. Assessing temporal performance, or computational complexity of the OSM method.

Response. For the purpose of performance analysis, we added the measurement of the execution time of clustering algorithms for the task of determining the dominant colors. The results are presented in the article.

Reviewer 3 Report

Comments and Suggestions for Authors

To ensure successful publication, the following reviewer comments have been carefully addressed:

  1. The ΔE₀₀ formula, representing color difference in human perception, might not be familiar to all readers. Therefore, a concise explanation has been provided in the abstract to clarify its relevance and usage.
  2. The introduction has been streamlined to succinctly highlight the research problem and its significance. The detailed literature review has been clearly separated into a dedicated "Literature Review" section (Section 2), effectively highlighting the gap between existing studies and the current research objectives.
  3. A clearer explanation has been provided regarding traditional image processing methods and deep learning techniques. The theoretical foundations, advantages, and limitations of each approach are systematically discussed. Empirical evidence supporting these discussions has been incorporated, including references such as "Effectiveness of image augmentation techniques on detection of building characteristics from street view images using deep learning." Additional relevant studies have also been cited to strengthen the empirical foundation.
  4. Figure 5 has been reinterpreted within the research context. Rather than describing general characteristics, the explanation now emphasizes how the sRGB color space projection onto the CaS chromaticity plane facilitates accurate color analysis crucial to the research methodology, highlighting its significance for consistent color representation and analysis in the context of the study.
  5. Numerical values presented in Table 5 have been consistently formatted to two decimal points to improve clarity and standardize presentation.
  6. Figures 8 and 9 have been distinctly described, clearly differentiating their contents and purposes. The description explicitly states their comparative analyses—Figure 8 focusing on dominant and prominent colors within the CIE ab chromatic plane and Figure 9 illustrating luminance (L) comparisons among various color clustering techniques such as K-means (KM), K-means in CIE Lab color space (KML), Fuzzy C-means (FCM), and the Orthogonal Saliency Mean (OSM) method developed specifically in this research.
  7. All abbreviations have been clearly defined upon first use, and repetitions of expanded abbreviations throughout the manuscript have been minimized for readability and consistency. A careful review was conducted to ensure that all terms are introduced properly and explained.

Author Response

We thank the reviewer for his comments that helped us improve the article.

1. The ΔE₀₀ formula, representing color difference in human perception, might not be familiar to all readers. Therefore, a concise explanation has been provided in the abstract to clarify its relevance and usage.

Response 1. Thank you for your comments, we have added an explanation of the formula ΔE₀₀ to the abstract.

2. The introduction has been streamlined to succinctly highlight the research problem and its significance. The detailed literature review has been clearly separated into a dedicated "Literature Review" section (Section 2), effectively highlighting the gap between existing studies and the current research objectives.

Response 2. The Introduction is divided into two parts, and the second part identifies the gap in existing research and justifies the need for new approaches.

3. A clearer explanation has been provided regarding traditional image processing methods and deep learning techniques. The theoretical foundations, advantages, and limitations of each approach are systematically discussed. Empirical evidence supporting these discussions has been incorporated, including references such as "Effectiveness of image augmentation techniques on detection of building characteristics from street view images using deep learning." Additional relevant studies have also been cited to strengthen the empirical foundation.

Response. Thank you for your comments. In our article, we refer to a literature source with a broad overview of traditional image palette processing methods, where the advantages and limitations are presented. Using this source [11] as a theoretical basis, we chose well-known methods to compare with the OSM method.

4. Figure 5 has been reinterpreted within the research context. Rather than describing general characteristics, the explanation now emphasizes how the sRGB color space projection onto the CaS chromaticity plane facilitates accurate color analysis crucial to the research methodology, highlighting its significance for consistent color representation and analysis in the context of the study.

Response. The comments have been taken into account in the updated version of the article.

5. Numerical values presented in Table 5 have been consistently formatted to two decimal points to improve clarity and standardize presentation.

Response. The comments have been taken into account in the updated version of the article.

6. Figures 8 and 9 have been distinctly described, clearly differentiating their contents and purposes. The description explicitly states their comparative analyses—Figure 8 focusing on dominant and prominent colors within the CIE ab chromatic plane and Figure 9 illustrating luminance (L) comparisons among various color clustering techniques such as K-means (KM), K-means in CIE Lab color space (KML), Fuzzy C-means (FCM), and the Orthogonal Saliency Mean (OSM) method developed specifically in this research.

Response. The comments have been taken into account in the updated version of the article.

7. All abbreviations have been clearly defined upon first use, and repetitions of expanded abbreviations throughout the manuscript have been minimized for readability and consistency. A careful review was conducted to ensure that all terms are introduced properly and explained.

Response. The comments have been taken into account in the updated version of the article.

 

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The author dealt with my comment. I have no further comment.

Reviewer 2 Report

Comments and Suggestions for Authors

No further comments need

Reviewer 3 Report

Comments and Suggestions for Authors

There is no further comments.

Back to TopTop