Exercise #11 Chroma key

For this exercise, we made a parody of an anime opening song using chromakeying. The output is as shown.

That is all.

Exercise #10 Color Tracking

For this exercise, we take a ROI on a webcam, detect the color, and track all colors that are the same or atleast within its threshold. Also, if the ROI should see black, The tracking is reset.

As its hard, no images are shown as reference.

The algorithm is as such:
1. First, make the ROI.
2. Save the ROI to get the colors on the region. Convert RGB to HSV and get the hue.
3. Compute the range and, using it, apply it to be tracked using inRange() function.
4. For tracking, code given can be used.

That is all.

Exercise #9 Background Subtraction

For this exercise, we are tasked to subtract the background of any clip. We used this.

First, we get all the frames of the video into a folder.
Then, taking the sequence of images, we convert them to gray values and get the median of a certain pixel from all frames. once the median is chosen, repeat for all pixels. Doing that, we will get the background of the image.
Next, using the background we got, all frames are again converted to its gray value.
Finally, subtracting the background, as long as its within the threshold, we will get the foreground for all the frames and compile.

That is all.

Exercise #8 Pseudo Coloring (False Coloring)

For this exercise, we were tasked to re-color the bodies of water on the following set of weather images.

After re-coloring, the outputs derived are:

We will use the first image for reference while doing this exercise.

First, get the grayscale and hsv counterparts of the image.

Using the function inRange(), get a binarized image where the bodies of water is white.

Using that on the original image, as long as it is white, recolor the the pixel to the desirable scale of your choosing. Depending on the intensity of the blues, recolor the image. The output is seen at the top.

That is all.

Exercise #4 Binarization and Canny Edge Detection.

For this exercise.

First, we binarized a bunch of images.
The following images are used for binarization.

For this set of images, we set thresholds for themselves for better binarization. As using one code for all is not an easy task for this one, we opted to hardcode it, making it so that each image use their own bits of the code.

The following are the output.

Next, for Adaptive thresholding, we used this sample image.

For its algorithm, we used the mean of the neighbors of the pixel of interest. Using it, we got this binarized image.

For detecting edges, we used a blur function built in in OpenCV. We also used Sobel and Canny Edge Detectors that are built in on the images.

That is all.

Exercise 1: Image Transition

The goal of the exercise is to transition one image to the other using a limited number of sequences.
The image used are:

That is all.

Exercise 3: Image Filtering

This exercise uses mean and median filters with different kernel configurations to get the image.


Kernel 1

Mean Filter

Median Filter

Not much can be seen from this image as they can both be used to filter the image. the higher the value of kernel size, the bigger the difference can be in terms of blurring the edges.



Kernel 2

Mean Filter

Median Filter

The median filter is lighter compared to mean filter. Also, the banana has more violet spots in the mean filter which correlates to the nature of what a mean is.



Kernel 3

Mean Filter

Median Filter

In the mean filter, you can still see the edges, but its darker. While in the median filter, the edges can be disregarded as you cannot see it.



Kernel 4

Mean Filter

Median Filter

The mean filter is same with kernel 3 but blurrier. The edges are scattered. The median filter is same with kernel 3.



Kernel 5

Mean Filter

Median Filter

The mean filter is buzzing with noises scattered to the places where this is much edges. As you can see, the surroundings of the banana is dark.
The median filter made the image turn negative.


That is all.