0 Replies - 670 Views - Last Post: 03 October 2020 - 05:09 PM Rate Topic: -----

#1 rouncer   User is offline

  • New D.I.C Head

Reputation: 1
  • View blog
  • Posts: 18
  • Joined: 09-May 18

I'm having trouble with scale invarient image keys (related to SIF

Posted 03 October 2020 - 05:09 PM

This is related to Scale invarient feature descriptors, is the most usual way to define it I think.

Do you guys know what i mean by key? In the sense im using them they are little patches of a photo that scroll along pictures, I use to match with.

You can give them invarience in the form of rotation and scale, as well as translation, by building the keys off the photo at different scales and rotations, (you detect what scale and what rotation you want before u build the key.)

The problem is, im not getting any matches for zooms, in the form me bobbing my head back and forth in front of the screen like an idiot. (Like most people look like when doing computer vision.)

It doesnt seem to arise at the same key at a different scale.

The form im using at the moment is so many bins x and y dividing the radius up, and then averaging all the pixels in each bin, rgb.
This doesnt seem to work - is there a better way to do it than that?

Is This A Good Question/Topic? 0
  • +

Page 1 of 1