Multi Features Data Clustering Using Novel Statistical Method with Application to Color Images

The efficiency and performance of the color image clustering algorithms are determined by various factors, including accuracy, data size, speed, and reliability (the absence of randomness in the results). Some applications, like microscopes analyzing images of biological objects or telescopes observing planetary motion prioritize accuracy over execution time. In contrast, surveillance cameras and moving object tracking prioritize speed and reliability over accuracy. This study introduces a novel algorithm that balances these four factors by clustering data with multiple features linked through specific relationships. The proposed algorithm has been practically applied to RGB color images. Traditional clustering methods, such as K-means, K-medoids, and Fuzzy C-means (FC-means), often prioritize arithmetic operations (e.g., means), which can deviate from true values, reducing calculation accuracy and resulting in outliers. Another limitation of traditional clustering methods is the irregularity in their outcomes, which reduces reliability due to the inherent randomness in selecting the initial point. These methods also perform poorly when handling large and multidimensional datasets. In contrast, the suggested clustering method employs a completely different approach. It identifies the Relationship Between Features Clustering (RBFC) of a specific object, represented across several matrices. It reduces their dimensions into a one-dimensional matrix representing the dissimilarities among such relations. The data is then clustered. Applying the RBFC method to a set of color images verified that the said method outperformed the aforementioned traditional methods insofar as clustering accuracy and processing time were involved. The suggested method also improves reliability by removing randomness and reliably producing color clusters in the same sequence throughout multiple runs. Moreover, the RBFC approach has shown remarkable performance in large datasets like medical and satellite images. The findings highlight that the RBFC approach is an efficient and effective clustering method for complex data scenarios and image processing applications.