Convex active contour model for target detection in synthetic aperture radar images

Fast and Accurate Target Detection Based on Multiscale Saliency and Active Contour Model for High-Resolution SAR Images

“…In fact, as we analyzed in [20], the aforementioned classical region-based [7]- [10] and hybrid [14]- [18] ACMs are mostly based on modifications of the geodesic active contours (GAC) [21], CV, and RSF (which is also a variant CV model with local region functions) models or are a combination of them. In order to improve the robustness of the aforementioned classical ACMs to multiplicative speckle noise in SAR image processing, a modified CV model was proposed in our early research [20] see (2), shown at the bottom of the page, where H(x) is the Heaviside function, i.e., if x ≥ 0, then H(x) = 1, and if x < 0, then H(x) = 0; p img represents the probability density function of the whole image, and it is a constant in practice.…”

“…In order to improve the robustness of the aforementioned classical ACMs to multiplicative speckle noise in SAR image processing, a modified CV model was proposed in our early research [20] see (2), shown at the bottom of the page, where H(x) is the Heaviside function, i.e., if x ≥ 0, then H(x) = 1, and if x < 0, then H(x) = 0; p img represents the probability density function of the whole image, and it is a constant in practice. In addition, p sin [z, φ(x)] and p out [z, φ(x)] are probability (2) density functions inside and outside the contour, respectively, and are defined as follows:…”

“…Therefore, as the distance decreases, the relativity between the pixels and their corresponding local region increases; thus, the aforementioned ACMs can divide an image into foreground and background. However, the Euclidean distance fails to measure the distance between two speckled SAR image patches (see a more detailed analysis in our early research [20]). Therefore, these classical ACMs are still sensitive to multiplicative speckle noise in SAR images.…”

“…1 This paper is organized as follows. Section II introduces a novel ACM, which is called the modified LGIF model and which was applied for effective target contour detection of SAR images in our earlier research [20]. In Section III, the edge indicator function in the modified LGIF model is improved with the ratio of exponentially weighted averages (ROEWA) operator [25], and an efficient split Bregman iteration approach is then utilized to solve the numerical problem of active contour propagation toward target boundaries.…”

Synthetic Aperture Radar Image Segmentation with Reaction Diffusion Level Set Evolution Equation in an Active Contour Model

“…As a result, the segmentation accuracy improved. Tu et al [25] proposed a convex ACM based on a ratio distance. The model was defined by a linear combination of the modified Chan-Vese model [15] and region-scalable fitting (RSF) energy model [18] in a level set framework.…”