Improved Image Segmentation Using Adaptive Thresholding and Morphological Refinement

Abstract

Image segmentation is a critical step in image analysis and computer vision, enabling the separation of meaningful regions or objects from background information. Accurate segmentation is essential for applications such as medical imaging, remote sensing, industrial inspection, and object recognition. Traditional global thresholding methods often fail in images with non-uniform illumination, noise, or complex backgrounds. To address these challenges, this paper proposes an improved image segmentation approach that combines adaptive thresholding with morphological refinement techniques. Adaptive thresholding dynamically determines local threshold values based on neighborhood intensity variations, allowing effective segmentation under varying lighting conditions. Morphological operations, including erosion, dilation, opening, and closing, are subsequently applied to refine object boundaries, remove noise, and fill small gaps. The proposed method is evaluated on standard benchmark image datasets and compared with conventional segmentation techniques. Experimental results demonstrate that the proposed approach achieves superior segmentation accuracy, improved boundary preservation, and enhanced robustness to noise. The findings confirm that integrating adaptive thresholding with morphological refinement provides an effective and computationally efficient solution for complex image segmentation tasks.