Boundary Detection of Objects in Digital Images Using Bit-Planes and Threshold Modified Canny Method

摘要

Two novel Canny-based boundary detection techniques are presented in this paper. Canny edge detection has gained popularity over the period due to its potential in edge detection. However, the edges detected by Canny are highly superfluous to extract the boundary of the objects in an image. The Modified Canny methods address this issue by modifying the parameter of Canny. The first method namely Threshold Modified Canny (MC-T) uses the Mean of the input image as threshold. MC-T is found to produce the boundaries even on the high-contrast images. The Second method, Bit-planes and Threshold Modified Canny (MC-BT) performs edge detection on the three intensity significant bit-planes using Mean of the input image as Threshold. This technique has also produced promising results in detecting the image boundary. The second method as it works only on three bit planes information of the input image, it reduces insignificant details and yields significant object boundaries. The result of the two proposed techniques, suitably finds place in object recognition, pattern recognition / matching etc. where boundary detection is an important component. These approaches are much promising in terms of clear boundary detection of an object, as boundary detection by conventional methods is very time consuming.