Snake Validation: A PCA-Based Outlier Detection Method

We utilize outlier detection by principal component analysis (PCA) as an effective step to automate snakes/active contours for object detection. The principle of our approach is straightforward: we allow snakes to evolve on a given image and classify them into desired object and non-object classes. To perform the classification, an annular image band around a snake is formed. The annular band is considered as a pattern image for PCA. Extensive experiments have been carried out on oil-sand and leukocyte images and the performance of the proposed method has been compared with two other automatic initialization and two gradient-based outlier detection techniques. Results show that the proposed algorithm improves the performance of automatic initialization techniques and validates snakes more accurately than other outlier detection methods, even when considerable object localization error is present.

I. Introduction

Research on snake or active contour [6]-based techniques has made them a mature interactive image segmentation tool. However, the techniques are yet to become fully automated in many applications. For complete automation, one needs the following three sequential steps: 1) snake initialization; 2) snake evolution; and 3) validation of the evolved snakes. Literature survey shows that much of the effort to date has been exerted on the first two steps, while the last step is practically ignored and left for the application to decide. The automated initialization techniques proposed to date mostly attempt to exploit the structure/shape of the objects (e.g., see [2] and [8]). In this paper we argue that a very crucial step for overall full automation of the snake is necessarily the validation step—often more important than the initialization step—when simple blind/random initializations are possible.