The corner detection algorithms described here are available for online testing over the Internet. Two testing options are provided. One can either select one of the standard shape images provided, or submit his own image and obtain the results. This gives a potential user a possibility to select an algorithm that fits his application.
A grayscale, 8 bit-per-pixel image file can be submitted. Most of the widespread image file formats are processed, e.g., TIFF, GIF, JPG, PCX, SUN raster, and many others -- those that are automatically recognised by the convert program of the ImageMagick package.
The image being submitted should contain one or more shapes of more or less uniform intensity viewed against uniform background. The image is automatically binarized using the (non-adaptive) threshold selection algorithm by Otsu. (If the input image is already binary, it is left intact.) After connected component analysis, small blobs are removed, while each of the large blobs is represented by a closed, chain-coded contour. Corners are then detected in each of the shapes.