In the exercise we, Eliza and I, were tasked to do the following:
- Convert an Image to Greyscale.
- Create Binary Image using Adaptive Thresholding.
- Apply a Canny function to a Binary Image.
- And then explain our approach on this, especially since there are many approaches to image processing. There isn’t only one way of doing things, feature multiplicity.
The first and third step were completed in class, while the second and fourth was left to our processing.
The original image is a 300 by 300 pixel image. But since we had more difficulty processing an image that small, since we couldn’t interpret properly whether we have achieved the certain effect or not, we resized the image into 650 by 650.
at the upper right is the resized image while the image at the upper left is its binary output. We put a transformation function inside the binarize function to separate the values from each other, before binarizing the values to be return as the focus point.
PROCESS OF BINARIZATION
But before actually binarizing and transforming the image, we first changed the image to greyscale, since the image can be humanly interpreted as an image full of text the best option is to changed the image into greyscale and then blur and sharpen the image enough that noise will no adversely effect the process of binarizing.
After, binarizing we still blurred and shapenned a little the binarized image since the initial output was like this:
Where the edge detected for each word was not that distinct from the other words. To make it clearer, the image went through another 3 processes to make it clearer for the Canny Detector to detect the edges.
The workload for this exercise was divided in such a way that initial preprocessing code was done during class our in a single computer, and then later separated to do tweaking of adding additional blur and sharpen functions.