Re: 'Re: Ups.

Wolfgang Müller wrote:

[snip]
>> BTW, I do think that not rescaling will cause serious problems, since
>> the feature extraction code is premised on 16x16 blocks at the lowest
>> level, and using images of different sizes will result in the same block
>> feature ID being used to refer to different image locations for
>> different images. IDs are assigned simply by counting.
>>     
>
> Yes, this is clear. To translate: all images of one collection are to be 
> resized to the same size. However, changing this size is not a problem 
> (within limits). Right?
>
>   
Yes, with rewritten feature extraction code, this should be fine. The 
current code in GIFT has '256' hard-coded into it though.

The best solution would probably be to rewrite the code so that it used 
a fixed depth quadtree, but tolerated non-square blocks. This would also 
have the advantage the texture distortions through the rescaling of 
non-square images would be avoided.

Regards and Happy Easter,

David