Yapay Zeka Hızlı Bakış & Çıkarımlar
AI InsightThe edge detection results are within acceptable measures and have responded to high performance and have the feature of to be applicable to large image types.
📌 3 Maddede Bu Makale
Abstract
Borders of objects and shadows in the image, reflections and lighting changes within objects are named as edge. The image features of the pixel with itself and its neighbors, play a significant role in detection of the edges. Automatic threshold edge detection algorithms on the similarity image obtained from color images have been proposed in this study. Firstly, the relation matrix based on the similarity feature between neighbor pixels is utilized and the color image is converted into twodimensional similarity image. In the second stage, histogram curve and fuzzy c-means method have been employed to obtain the automatic threshold value. Threshold values obtained by virtue of these two methods have been applied to similarity images obtained separately by Linear, Exponential and Gaussian functions. Visual results have been utilized for the performance evaluations of the two algorithms. Thin edges have been created in the histogram-based edge detection algorithm while distinct and thick edges have been created in the fuzzy c-means algorithm. Clear and distinct edges have been created in linear and exponential functions. The results of the other two methods have been achieved in the Gaussian function, through utilization of the low D coefficient. The edge detection results are within acceptable measures and have responded to high performance and have the feature of to be applicable to large image types.