E186 Project 2 (due Feb. 21)

1.

Image ``/e186/imagedata/noisy_taxi'' is an image of a street scene of a city on a newly discovered planet in the outer space, which was transmitted back to Earth from the Hubble telescope. As the communication is very noisy, the image is badly contaminated by a particular type of noise. Try the following algorithms to reduce as much noisy as possible while still maintaining as much sharpness of the image as possible.

(a)

Spatial low-pass filtering (using the best kernel of your choice).

(b)

Median filtering (using the best window shape/size).

(c)

Nagao method (try a few different window sizes to find the best)

Compare the pros and cons (effectiveness, computation time, etc.) of these algorithms, and explain why. Print the image after the noise is reduced for each of the algorithms. Submit them as images 1, 2, and 3.

2.

Equalize the histogram of the best resulting image you got from the first part. Submit this image with equalized histogram as image 4. Also submit the histograms of the images before and after equalization as images 5, 6.

3.

Modify the histogram of the same image as used in the second part so that it linearly increases from grey level 0 to 127, and then linearly decreases from grey level 128 to 255. Submit the image after the processing as image 7 and its histogram as image 8.

4.

Detect the edges in images ``/e186/imagedata/rubic''. The output image should show the edges in binary (e.g., black background and white edges). You may need several separate steps for this task, such as (i) preprocessing the image (linear stretching, etc.), (ii) edge detection and (iii) thresholding. Use each of the following methods:

(a)

Gradient Operators

Try each of the gradient kernels:

i.

Roberts operators

ii.

Prewitt operators (try both 3x3 and 4x4)

iii.

Sobel operators

(b)

Compass Operators:

Use compass operators of all 8 directions and then find the maximum as an indicator of the local change.

Submit these four results as images 9, 10, 11, and 12.

5.

The goal here is to extract the complete boundaries and edges of the objects in the image rubic. As this goal might not be easily reachable, this is really an open-ended problem. Now try your best to combine different methods or to come up with your own. Submit your best result as image 13.