A multilevel thresholding algorithm using Electromagnetism Optimization

Segmentation is one of the most important tasks in image processing. It consist in classify the pixels into two or more groups depending on their intensity levels and a threshold value. The quality of the segmentation depends on the method applied to select the threshold. The use of the classical implementations for multilevel thresholding is computationally expensive since they exhaustively search the best values to optimize the objective function. Under such conditions, the use of optimization evolutionary approaches has been extended. The Electromagnetism Like algorithm (EMO) is an evolutionary method which mimics the attraction repulsion mechanism among charges to evolve the members of a population. Different to other algorithms, EMO exhibits interesting search capabilities whereas maintains a low computational overhead. In this paper, a multilevel thresholding (MT) algorithm based on the EMO is introduced. The approach combines the good search capabilities of EMO algorithm with objective functions proposed by the popular MT methods of Otsu and Kapur. The algorithm takes random samples from a feasible search space inside the image histogram. Such samples build each particle in the EMO context whereas its quality is evaluated considering the objective that is function employed by the Otsu or Kapur method. Guided by these objective values the set of candidate solutions are evolved through the EMO operators until an optimal solution is found. The approach generates a multilevel segmentation algorithm which can effectively identify the threshold values of a digital image in a reduced number of iterations. Experimental results show performance evidence of the implementation of EMO for digital image segmentation.

💡 Research Summary

The paper presents a novel multilevel thresholding (MT) method for image segmentation that leverages the Electromagnetism‑Like Optimization (EMO) algorithm. Traditional MT techniques, such as those based on Otsu’s between‑class variance or Kapur’s entropy, require exhaustive search when the number of thresholds increases, leading to prohibitive computational costs. To overcome this limitation, the authors embed EMO—a population‑based meta‑heuristic that mimics the attraction‑repulsion dynamics of charged particles—into the threshold selection process.

In the proposed framework, the image histogram defines the feasible search space. Random samples are drawn from this space to construct candidate solutions (particles), each particle being an ordered vector of k thresholds (k is user‑specified). The quality of a particle is evaluated using either Otsu’s variance‑maximization criterion or Kapur’s entropy‑maximization criterion, allowing the algorithm to be tailored to contrast‑or information‑preserving segmentation goals. EMO then assigns a charge to each particle proportional to its fitness, computes pairwise forces (attractive for higher‑fitness particles, repulsive for lower‑fitness ones), and moves particles along the resultant force vectors. Boundary handling (reflection or clipping) ensures that thresholds remain within the histogram limits. This iterative process continues until convergence or a preset iteration limit is reached.

Experimental validation employed standard grayscale images (Lena, Cameraman, Baboon) and noisy variants, testing 2 to 5 thresholds. Performance metrics included Peak Signal‑to‑Noise Ratio (PSNR), Structural Similarity Index (SSIM), and execution time. Results demonstrate that both EMO‑Otsu and EMO‑Kapur achieve segmentation quality comparable to or better than exhaustive‑search baselines while reducing computational time by an order of magnitude. The advantage is especially pronounced for higher‑dimensional threshold sets (four or five thresholds), where EMO maintains robust global search capability and avoids premature convergence to local optima, even under significant noise.

Key contributions of the work are: (1) introducing EMO as an efficient optimizer for MT, dramatically lowering the combinatorial explosion inherent in exhaustive methods; (2) integrating two widely used objective functions (Otsu and Kapur) within a single, flexible framework; (3) employing histogram‑based random sampling to shrink the candidate space without sacrificing diversity.

The authors acknowledge several limitations. Convergence speed can be sensitive to the initial particle distribution, and EMO’s hyper‑parameters (population size, step‑size factor) require empirical tuning for different image characteristics. Moreover, highly sparse or irregular histograms may cause instability in charge calculation. Future research directions include adaptive parameter control, hybridization with other meta‑heuristics, and real‑time implementation for video streams.

In summary, the EMO‑driven multilevel thresholding algorithm offers a compelling balance of computational efficiency and segmentation quality, positioning it as a practical alternative for a broad range of image analysis and computer‑vision applications.