TextPoint, arecognition- recall CaRP scheme with an alphabet of characters. Fig. 5 showssome invariant points of letter “A”, which provides a strong cue to locate itsinvariant points memorize. If the distance to the closest boundary of the objectexceeds a threshold that point is known as the internal point of an object.
ForTextPoints, to form a set of clickable points, a set of internal points ofcharacters is selected. Here there is a chance of overlap between clickablepoint and neighboring character and their tolerance ranges also occluded on theimage generated by the underlying Captcha engine. In determining clickablepoints to avoid overlap on CaRP images, the distance between any pair ofclickable points in a character must exceed a threshold by this both can beseparated. Consider an example, if the center of a stroke segment in onecharacter is selected, the user cannot select the center of a similar strokesegment in another character.
To avoid this, the user must select a differentpoint from the stroke segment. By this user can ensure that a clickable pointis context-dependent: depending on the character that the point lies in thesame type of structured point may or may not be the clickable point. Eventhough the clickable points are known for each character, character recognitionis necessary for locating clickable points on a TextPoints image. These things cannot be achieved by bots