Abstract:Image segmentation is a crucial process in the field of computer vision, and deformable models for image segmentation have demonstrated remarkable efficacy among various techniques. These models have gained significant attention due to their ability to generate closed and smooth contours of target boundaries. Nevertheless, the conventional deformable models exhibit undesirable performance when confronted with complex images having severe inhomogeneous intensity and speckle noise. To address these challenges, we present a new coupled deformable model (CDM) that utilizes the level set framework by incorporating the idea of inhomogeneity correction to achieve precise image segmentation and correction. Furthermore, we introduce a despeckling equation that effectively eradicates the speckle noise while preserving important image characteristics. This equation is derived from the Rayleigh distribution to tackle the issue of speckle noise that usually appears in ultrasound images. In addition, a separate equation is employed to enhance the edge quality, thereby improving segmentation outcomes. The proposed framework successfully eliminates the tedious re-initialization and addresses the issue of numerical instability. The well-posedness of the present model is rigorously established by utilizing Schauder's fixed point theorem. Extensive experiments are conducted on a diverse range of synthetic and real images contaminated by severe intensity inhomogeneity and speckle noise. The comprehensive qualitative and quantitative analysis demonstrates the superiority of the present model over classical and recent representative deformable models. Moreover, the proposed model exhibits robustness against various levels of noise and severe intensity inhomogeneity, while remaining insensitive to the initial contour placement.

Segmentation Through Point-based Multiscale Deformable Model

Neighborhood Aided Implicit Active Contours

Local Weak Form Geometric Active Contours For Medical Image Segmentation

Point-Based Geometric Deformable Models for Medical Image Segmentation

Neighbor-constrained Active Contours Without Edges.

A Medical Image Segmentation Algorithm Based on Improved DDCM Deformable Model

Dpnet: End-to-end Aerial Image Segmentation Via Deformable Point Network.

Biomechanical Model Constrained Image Segmentation Framework

ECDM: Enhanced Edge Based Coupled Deformable Model for Image Segmentation in the Presence of Speckle Noise and Severe Intensity Inhomogeneity

Using Surface Variability Characteristics for Segmentation of Deformable 3d Objects with Application to Piecewise Statistical Deformable Model

A Multiscale Circular Geodesic Model for Image Segmentation

MODELING SEGMENTATION VIA GEOMETRIC DEFORMABLE REGULARIZERS, PDE AND LEVEL SETS IN STILL AND MOTION IMAGERY: A REVISIT

PointMS: Semantic Segmentation for Point Cloud Based on Multi-scale Directional Convolution

Detail-Preserving local editing for point-sampled geometry

Deformable-Model based textured object segmentation

Multi-Scale B-Spline Level Set Segmentation Based on Gaussian Kernel Equalization

A Deformable Local Image Descriptor

An Improved Geometric Deformable Model for Color Image Segmentation

A LOCAL NONPARAMETRIC MODEL FOR SIMULTANEOUS IMAGE SEGMENTATION AND ADAPTIVE SMOOTHING

MULTIRESOLUTION ADAPTIVE IMAGE SEGMENTATION BASED ON DEFORMABLE MODELS

Dual Adaptive Transformations for Weakly Supervised Point Cloud Segmentation