Local Image Analysis of Malaysian Herbs Leaves Using Canny Edge Detection Algorithm

Machine vision helps a lot with the latest recognition technology. To get the best recognition results, the initial phase of image processing should be done as best as possible. This phase involves the production of an image map of the resulting image using edge detection. The Canny method is often used because of its performance of producing meticulous strong edges but this method is sensitive to changes in image intensity when involving complex images such as image leaves and results in a lot of noise at the edges of the resulting image. This is because in this method the threshold value is selected empirically on the image globally. In this study, local image analysis will be discussed along with its impact on the resulting image edge results. In addition, a set of data from herbal leaves in Malaysia has also been produced by containing ground truth images for each herbal image available. The results from this study found that the locally analysing image approach has outperform the global approach of the conventional Canny method. Findings from this study may help the identification system in the future.