Total Variation-based Active Contour Model for Medical Images

Abstract

The Active Contour Model (ACM) is a mathematical model in image processing that is commonly utilized to partition or segment an image into specific objects. The segmentation method in region-based ACM can be categorized into two classes: global ACM and selective ACM Selective ACM isolates a specific target item from an input image, which is more advantageous than the global ACM due to its proven use, particularly in medical image analysis. However, the selective ACM appears to produce poor outcomes when segmenting an image with uneven (inhomogeneous) intensity. Additionally, the current selective ACM that uses the Gaussian function as a regularizer generates a non-smooth segmentation curve, especially for images containing noise. This study introduces a new selective ACM that is designed to segment medical images with inhomogeneous intensity levels. The model incorporates a Total Variation term as a regularizer, distance function, and local image fitting concepts. The Euler-Lagrange (EL) equation was given to solve the suggested model, which is approximately 5% more accurate with a processing time that is around three times faster than the existing model, as shown by numerical testing. The suggested mathematical model can be advantageous for the image analysis community, particularly in the medical industry, to automatically segment a specific object in a medical image.