Algorithm of Color Fusion for Multispectral Grayscale Images based on K-Means Clustering of Sample Image
Abstract
Introduction: Multispectral aviation systems of technical vision find extensive application in optoelectronic investigation, in providing better awareness of mobile equipment crews, or in control under difficult weather conditions. Information from monochrome sensors with different spectral characteristics is pooled using special fusion algorithms. However, the existing fusion algorithms are usually sensitive to noise in the video data due to their limited dynamic ranges of light and color, errors in the calibration characteristics and in the temporal bindings of the video stream. Purpose: An algorithm should be developed which would provide a structurally stable solution for the problem of multispectral image data fusion on the set of possible photometric situations. Results: Fusion methods have been reviewed for grayscale television and thermal images, forming a grayscale or false-color result. The global methods has been discussed for transferring the color from a sample image. An algorithm has been developed for computing the color characteristics of a grayscale complexed multi-spectral image based on a sample image clustering. This algorithm comprises the following: grayscale fusion to form the brightness components of the final image; converting the sample image from RGB color space to YIQ space; color segmentation of the sample image in YIQ space using the K-means clustering algorithm; search for each pixel of the initial image looking for the cluster with the nearest brightness and assigning the color information of the cluster to the chromatic components of the pixel; converting the obtained complexed image from YIQ color space to RGB space. Concrete examples of image processing in MatLab prove the effectiveness of the proposed algorithm as compared to the existing global fusion methods. Practical relevance: Analysis of normalized histograms has shown that the proposed algorithm of calculating the color characteristics provides a higher quality of determining color information as compared to «transferring» the color from a sample image.Published
2017-12-20
How to Cite
Neretina, V., & Efanov, V. (2017). Algorithm of Color Fusion for Multispectral Grayscale Images based on K-Means Clustering of Sample Image. Information and Control Systems, (6), 15-23. https://doi.org/10.15217/issn1684-8853.2017.6.15
Issue
Section
Information processing and control