Tracing the contour of an object in an image is an important problem in computer vision. This paper presents a new contour detection algorithm using an adaptive vision framework. The proposed method is different from conventional algorithms in several ways. First, we introduce a new adaptive tracking algorithm called LR-traversing. LRtraversing is unique as it progressively adapts to the thickness of an edge while tracking the contour of an object with variable sharpness. Secondly, the method employs adaptive selection process that can optimally extract features based on an error metric. By utilizing this flexible run-time technique our method can detect and track object contours in realtime. Experiments demonstrate that the method is significantly faster than other algorithms that can achieve similar result.