A new algorithm called an anchored discrete convolution has been developed. The method enables the determination of the length of objects in digital images from their detected contour. The algorithm ...is based on discrete convolution by using different convolution masks and points anchored within the pixels. The anchoring prevents the point shifting in the pixel during the calculation and contributes to smoothing capabilities of the algorithm on the measured data. In this paper a Gaussian convolution mask is used on the experimental curves to evaluate the presented algorithm as a measuring tool.
The paper presents a new method introducing an anchored discrete convolution for calculating the length of a digital curve. The method is based on discrete convolution by using convolution masks and ...point anchoring in the pixel. The use of ordinary convolution distorts the curve shape and gives large errors in length calculation. The advantage of anchoring is that it limits the point shifting into the pixel during the calculation of the curve length. The method is applied to an analytical arc and various calculations are performed. In addition different methods from the literature were compared and a real sample was tested.
A digital image could be used as a source of data for various dimensional measurements. After acquiring the digital image of the measured object, the object contour is determined. The contour is ...composed of a sequence of distinct pixels which contain centre points. This is a representation of a digital curve or digital arc from which the object length or the perimeter could be determined. In order to calculate the approximate length of the object, we introduce a new algorithm called an anchored discrete convolution. The algorithm is based on a discrete convolution by using different convolution masks and points anchored in the pixels. Unfortunately, the use of an ordinary convolution distorts the curve shape and introduces large errors in length calculation. In our case the anchoring limits the point shifting into the pixel area during the calculation of the curve length. In this paper different convolution masks are used on an experimental digital curve in order to evaluate the presented algorithm as a measuring tool. In addition, different real objects were optically scanned and the perimeters of the objects were determined from the acquired images.
PDF