## Problem with given sift.m

Moderator: Computer Vision

leviathan
Computerversteher
Beiträge: 307
Registriert: 30. Jul 2008 14:26
Wohnort: Darmstadt
Kontaktdaten:

### Problem with given sift.m

We are trying to solve the current homework assignment using the sift, harris and find2chi functions provided in the email (on July 4th). Doing that, we have run into the following problem:

Code: Alles auswählen

??? Subscript indices must either be real positive integers or logicals.

Error in ==> sift at 36
dxb = dx(py(i)-7:py(i)+8, px(i)-7:px(i)+8);
This happens when we apply the given SIFT function, using the harris interest points also generated by the given harris.m functions. What happens here (I think), is that harris finds interest points on the image border, and trying to obtain the 16x16px window around them results in negative pixel coordinates (subscript indices). In our own solution for the previous exercise, we had a workaround for this problem; however, as a result, our SIFT function worked much slower than the reference implementation and is thus unsuitable for the current assignment (processing all interest points from all the training images would take forever).

How do we work around this problem?..
Ein Programmierer hat immer eine Lösung. Die passt nur nicht immer zum Problem.

Hiwi für Weiterentwicklung des Lernportals (Moodle).

tanne
Endlosschleifenbastler
Beiträge: 162
Registriert: 30. Sep 2008 16:05

### Re: Problem with given sift.m

just drop all points beyond a certain treshold at the borders of the image, just like in the last exercise.

qgao
Moderator
Beiträge: 33
Registriert: 7. Apr 2009 11:38

### Re: Problem with given sift.m

tanne hat geschrieben:just drop all points beyond a certain treshold at the borders of the image, just like in the last exercise.
Yes, exactly.