Subpixel edge estimation using geometrical edge models with noise miniaturization

The significant disadvantage for traditional contour representation is as the resolution is reduced the effort of undersampling is proportional enlarged. The goal of this study is to improve edge detection results, especially for those corner points in low resolution. This study describes a method, which is based on 4-connected pixel-wise linearization, for finding contours from low resolution video images. This allows a more accurate inspection and identification of objects from image data. In practice, geometrical models are used to manipulate this linearization. A method is employed for examining the corner points as well