ICS2412 Digital Image Processing For parts 1 and 2,and 3 use the “lenna” and “peppers” images; note that each one of them is a 256 x 256, 8 bits/pixel image. 1. Image Sampling Write a MATLAB program to change the spatial resolution to 128 x 128, 64 x 64, and 32 x 32 pixels. Resize the images back to the original size 256 x 256 images and print them for comparison purposes. Explain how your algorithm works. 2. Image Quantization Write a MATLAB program that would reduce the number of gray levels in a PGM image from 256 to: (i) 128, (ii) 32, (iii) 8, and (iv) 2. Explain how your algorithm works. 3. 3. Histogram Equalization (a) Write a MATLAB computer program to compute the histogram of an image. (b) Implement the histogram equalization technique. It is suggest that you debug your algorithm using a small “test” image (e.g., 5 x 5) to make sure that it works correctly. (c) Perform histogram equalization on the “lenna” and “peppers” images. (d) Show and discuss your results (i.e., original images/histograms, output images/histograms). Laboratory Write-up For each programming assignment, you are to turn in a brief report The report is very important in determining score for the programming assignment. Be well organized, type your reports, and include figure captions with a brief description for all the figures included in your report. 1 Fig. 1 “Lenna” image Fig. 2 “Peppers” image 2 A) • • • Write a Matlab script code to Read a simple RGB image "football.jpg". Create a separate image for each color planes (red, green and blue) of the image. Display each color plane image separately, and also display the original image. B) C) Geometric Transformation 1. Define a function which will rotate an image about its center pixel by a given angle. Test your function on an image of your choice. See Figure 1. 2. Figure 2 shows an image take by a camera pointed at a cone shaped mirror. Define a function which computes the geometric correction yielding a 360 panorama. See Figure 3. 3 4 E) Overview: In this assignment, you will apply various image restoration techniques to restore an image corrupted by noise. Assignment specifics: Part A: Search the NET for a file NoisyImg.bmp (or any other Noisy image), which is a 256-level gray image of a truck in a desert which has been corrupted by noise. First, apply median filtering to the given image and save the restored image as Median.bmp. Next, apply adaptive Wiener filtering to the given image and save the restored image as Wiener.bmp. Implement adaptive median filtering and apply it to the given image, and save the restored image as AdaptiveMedian.bmp. Devise your own technique by combining two or more of the previous techniques to improve the quality of the restoration. Compare and contrast your result with the previous three, and save the restored image as ResultA.bmp. NB. The Matlab imnoise() function can be used to add noise to images. F) Obtain the images Suzi1.bin and ct scan.bin from the Internet. Each image has 256 256 pixels and each pixel has 8 bits. In this assignment you will perform object extraction (target extraction) by using simple thresholding followed by connected components labeling (blob coloring) with minor region removal. This is a special case of two classical image processing problems known as image segmentation and classification. Throughout the assignment, including the printing of your results, use a value of 255 (Hex 0xFF) for LOGIC ONE and a value of zero (Hex 0x00) for LOGIC ZERO. Objectives: 1. Suzi1: the first objective is to produce a binary image J that is LOGIC ONE at pixels contained in the “girl” object of the original image and that is LOGIC ZERO at pixels contained in the background of the original image. The second objective is to produce a grayscale image K of the segmented “girl” object. At pixels where J is LOGIC ONE, K should be equal to the original Suzi1 image. At pixels where J is LOGIC ZERO, K should be 255. 2. ct scan: the first objective is to produce a binary image J that is LOGIC ONE at pixels contained in the “torso section” object of the original image and that is LOGIC ZERO at pixels contained in the background of the original image. The second objective is to produce a grayscale image K of the segmented “torso section” object. This should be done exactly the same way it was for the Suzi1 image. 5