In this assignment you will run a variety of edge detectors, and explore their operation on some sample images. Recall that most edge detectors follow a similar sequence of operations:
- Smooth the image
- Find the gradient by convolving with Gx and Gy kernels
- Select pixels where the gradient magnitude is over a chosen threshold. These are the edge pixels.
Each of these steps may be carried out in a number of different ways. The main differences between many of the approaches we have looked at lie in the choice of gradient kernels Gx and Gy. However, different results can also be obtained by varying the smoothing radius in the first step or the threshold in the last step.
Exploration
We will use Matlab's built-in edge detection function (edge)to explore the effect of different parameters. Read the documentation so that you know how to specify parameter values. Then try different edge detectors with different parameter settings on the images below, playing with the smoothing and thresholds until you find a level that you like. You should experiment with all the options that edge provides. Note that you will probably have ro convert color images using rgb2gray.
Questions
Here are some questions to consider when you are done:
- Which edge detectors give similar results? Which ones are clearly different? Can you comment on why this might be?
- Report your favorite edge detector and parameter settings for each image, and show the results. Explain why you chose these results over others. How do the wrench edge results compare to the outline image produced for the wrench image on the last homework?
- As a result of this exercise, can you identify any areas where the edge detectors you have tried are not giving you the results you would like? If you had to make a line drawing of each of these images, would your drawing look like the edge-detected result? Identify any major differences.