Stories
Slash Boxes
Comments

SoylentNews is people

SoylentNews is powered by your submissions, so send in your scoop. Only 10 submissions in the queue.
posted by Fnord666 on Monday September 18 2017, @12:08PM   Printer-friendly
from the whiteboards-never-looked-so-good dept.

Submitted via IRC for SoyCow5743

Microsoft's Pix sets itself apart from other camera apps by using the power of artificial intelligence to correct your photos, learning new tricks over time. It can do things like add artistic flair to your images, turn photos shot in a row into "Live Images," or just making sure the people in your photos look great. This week, the app got a new update out that adds yet another AI trick to the pile: The ability to capture whiteboards and turn them into useful images.

So, for example, if you're at an important meeting, you can use Pix to take a photo of a diagram on the whiteboard to remember it later. The Pix app will then sharpen the focus, ramp up the color and tone, crop out the background and realign the image appropriately so that the diagram is shown straight-on.

According to Microsoft:

The updated app automatically detects whiteboards, documents and business cards in real time and intelligently adjusts camera settings for these types of photos. Once the shutter clicks, the app uses AI to improve the image, such as cropping edges, boosting color and tone, sharpening focus and tweaking the angle to render the image in a straight-on perspective.

Source: https://www.engadget.com/2017/09/15/microsoft-pix-uses-ai-to-make-whiteboard-photos-useable-images/


Original Submission

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 1) by pTamok on Monday September 18 2017, @07:59PM

    by pTamok (3042) on Monday September 18 2017, @07:59PM (#569891)

    This is the sort of stuff you learn how to do in an image processing course, and doesn't really sound like "AI" has much, if anything, to do with it.

    If the smartphone/camera back-end processor is doing it without human intervention, then Artificial Intelligence may well be involved.

    1) Identify the corners of the trapezoid. Generate virtual ones if they are outside the field of view
    2) Apply a geometric transform to transform the trapezoid into a regular quadrilateral, then crop
    3) Sharpen up the line boundaries, making sure continuous lines remain continuous
    4) Re-map the palette with some intelligent cut-offs to make the whiteboard white, and enhance the colours of the markers.

    All of the above is easily done in an image processing workflow, but choosing the right parameters without human intervention is where an AI might be useful.