All Key Points and supplements would be attached FOR CONSIDERATION

Part I: Filter basics

Sharpen rain.jpeg and eliminate rain drops in it by median filter.

Task 1: Implement your sharpening filter and perform it on rain.jpeg. Save your results as sharpened.jpg. To get an obvious result, you may chose a relatively large ?. (2 points)

Task 2: Implement your own median filter and perform it on rain.jpeg to eliminate rain drops. Save results to derained.jpg. To get a good deraining result, the filter size should be as small as possible but large enough to eliminate rain drops. (2 points)

Part II: Canny edge detector

Implement Canny edge detector to detect edges in road.jpeg. Your program should contain following steps:

1. Convert colour image into grayscale image, and save as gray.jpg. (1 point)
2. Choose a proper sigma to perform Gaussian smoothing, save the result as gauss.jpg. Discuss how sigma affects final edge detection results. (2 points)
3. Apply sobel operator, save the x-gradient Gx, y-gradient Gy and magnitude G as G_x.jpg, G_y.jpg and G.jpg. (3 points)
4. Non-maximum value suppression. Save suppression result as supress.jpg. (3 points)
5. Hysteresis thresholding. Choose two proper threshold – low and high, then binarize your suppression result via Hysteresis thresholding. Save binarization result by lower threshold as edgemap_low.jpg, result by high threshold as edgemap_high.jpg. Save the final result by Hysteresis thresholding as edgemap.jpg. Discuss how different thresholds affect final edgemap. (2 points, one point for edgemap_low.jpg and edgemap_high.jpg, one for edgemap.jpg)

Bonus: Hough transform

This part follows part II Canny edge detector to recognize straight lines in edgemap.jpg.

1. Hough Transform. Save the “voting” result as hough.jpg. (2 points)
2. Find out several most possible straight lines. Draw them on original image, then save as detection_result.jpg. To draw lines on image, see this example. (1 points)