This study illustrates the performance analysis of the most
commonly used edge detection techniques including Canny, Sobel and Prewitt,
highlighting their advantages and disadvantages with respect to different types
of datasets. One of the most important stages
in image processing to find and detect discontinuities in intensity change is
edge detection. It is a useful tool for identifying various aspects of a
picture, including shape, contrast, color, scene analysis, and image
segmentation. The technique is very important to recognize all the edges
accurately. It helps in object recognition, pattern recognition, medical image
processing, motion analysis etc. There are many edge detection operators
available in image processing. After
analyzing various parameters like Accuracy, Mean Square Error (MSE), Peak
Signal to Noise Ratio (PSNR), Edge Detection Processing Time and Qualitative
Human Visual Perception on two diverse type of datasets, varied results are
found with respect to the techniques used. Among them, the most accurate and
fast computed edge detection technique which gives better results on both type
of datasets is concluded. Although the Sobel edge detection technique gives
relatively poor result and weak performance of detection of edges, however it
can be modified and further improved with respect to future work. The entire
analyzing process was done under Scilab software. Canny works well also but it
can be used for detecting very thin edges with the disadvantage of it cannot
detect object very precisely because of detecting small amount of intensity
variation. Sobel gives a very bad performance for small objects because it is
used for detecting thick edges so sobel is best fit for detecting satellite
images of large geographical area images. Future work can also be done for
video edge detection and an improved sobel edge detection technique can be
proposed which can detect thin edges also to overcome the disadvantage of
limitation of geographical area. In future work the platform of comparison can
also be change.
Author(s) Details:
Rajshree Kumari,
Department of Computer Engineering, G.B. Pant University of
Agriculture and Technology, Pantnagar, Uttarakhand, India.
Divyanshu
Chandra,
M.C.A.
Programme, G.B. Pant University of Agriculture and Technology, Pantnagar,
Uttarakhand, India.
Please see the link here: https://stm.bookpi.org/CPSTR-V8/article/view/14054
