distance-from-edge

Computer vision tool that finds the distance of a point from an edge using sobel edge detection and euclidean distance calculations.

For the caMicroscope GSoC 2021 Project Choice [2] Pathology Smartpens Code Challenge

Code Challenge: Create a page or tool which performs edge detection on a given image and, given a point, returns the distance from that point to the closest edge.

Personal Background Knowledge

For background, I took the course CSE 455 Computer Vision at the University of Washington, and in this course I implemented my own miniature computer vision library. One of the filters I implemented was the sobel operator for edge detection.

For this challenge, I added an additional function to my source code called get_smallest_dist_from_edge, which you may find in dist-from-edge/src/hw2/filter_image.c

I added this additional get_smallest_dist_from_edge on March 14, 2021.

I generated the example images with dist_from_edge.py

Test Image

Here is the image I used in my examples:

In the test image results shown below, the sobel edge detection filter has been applied. The test points are denoted with the smaller + and the closest edge detected is denoted with a larger cross.

User Interface

If you would like to try the tool out for yourself, make sure you have flask and flask-cors installed.

• pip install flask

• pip install -U flask-cors

Navigate to /dist-from-edge and run the backend, which will allow the page to communicate with the Python library to call the function.

• export FLASK_APP=distedge.py

• python -m flask run

Now, you may open the page /dist-from-edge/index.html in your browser.

Page Instructions: click on a point anywhere on the test image. There will be an alert which notifies the minimum distance from the closest edge, and the closest edge from the point you clicked will be shown in the result image.

Test 1 (point on beam)

Test Point: (x: 20, y: 100) denoted by small +

Closest Distance: 13.000000

Best Edge: (x: 8, y: 105) denoted by large +

Test 2 (point on dog)

Test Point: (x: 160, y: 400) denoted by small +

Closest Distance: 21.540659

Best Edge: (x: 140, y: 392) denoted by large +

Test 3 (point on car)

Test Point: (x: 510, y: 120) denoted by small +

Closest Distance: 7.810250

Best Edge: (x: 515, y: 126) denoted by large +

Next Steps

• Test more points on current test image.
• Add noise reduction.
• Test greater variety of images.