123
-=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- (c) WidthPadding Industries 1987 0|648|0 -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=- -=+=-
Socoder -> Blitz -> Image rotation?

Tue, 30 Nov 2010, 10:57
spinal
I am trying to rotate an image (yes I know there is a rotate image function, I'm not going to use it). Anyway, the simple way is to do the following --



However, it does leave gaps in the output image. I read somewhere that image rotation should be done backwards, as in, cycle through the output image and work out which pixel should be there, or something like that. But I have yet to find any info on how to do this.
Does anyone know how this is done?

-=-=-
Check out my excellent homepage!
Wed, 01 Dec 2010, 02:46
Jayenkai
It's quite awkward to think of this backwards, when you're working with pixels of a rotated rectangle.
Instead, simplify it.
You need to fill surrounding pixels of each pixel, so just do that.
Find each rx/ry, then go backwards for each pixel that's around it.



-=-=-
''Load, Next List!''
Wed, 01 Dec 2010, 09:22
Andy_A
A quick and dirty way to fill in the gaps is to plot a 2x2 rect instead of a single pixel.

The drawback is that your image may be slightly distorted due to the way the rects overlap.


Wed, 01 Dec 2010, 12:34
spinal
OK, I tweaked Jay's code a little, removed the initial rotation and instead worked completely backwards.

I assume though, that there is a simple way to find the min and max x and y values from checking the rotation of the image corners first though, I just guessed



-=-=-
Check out my excellent homepage!
Wed, 15 Dec 2010, 16:51
Teasy
That looks neat ^^

For finding the maximum image rotation corners,
I think you're looking for something like this, Spinal



To see it in action,
in a slightly modified program:



There is also room for more optimization ^^

Sat, 18 Dec 2010, 13:39
dna
Jay, In your code you use :

Graphics 1024,768,0,2
Graphics 320,240,0,3

What does that do? Why is that in there twice?

@Teasy: That's real smoothe. I can see where BB is slow when doing this.

Does BMax make a difference as far as the speed of the rot?


-=-=-
DNA
Sat, 18 Dec 2010, 15:06
Jayenkai
,2 says "Make the window THIS big"
,3 says "Make this a scalable window."

With a scalable window, you can Maximise it if you wish, but everything ends up looking a bit blurry and stretchy.
It's a nice thing to have, but ,2 will always be clearer.

However..
,3 doesn't set the window's default size.
So, by setting a LARGE ,2 (and making the window be 1024x768) and THEN setting a smaller ,3, you end up with the 320x240 graphics being rescaled up to the 1024x768 size.

Basically, whereas we're drawing things to only a 320x240 area, they're actually being displayed 3 times bigger on the screen.

Otherwise, you'd be looking at a teensy tiny window!

-=-=-
''Load, Next List!''
Sat, 18 Dec 2010, 16:03
CodersRule

I never knew you could do that.
I always wondered how to make a resizable window that big by default.
Sat, 25 Dec 2010, 06:31
Teasy

dna:
if you optimize the code way more,
it should be smooth and superfast in blitz2d even

i still have some of blitzcoder's code laying around
from optimization/math wizards ^^
which does exactly that!

i am not familiar
with blitzmax' graphics interface/library
(opengl) benchmarks,
but i imagine that it usually depends
on (the wishes of) the coder
whether something is fast or not

taking into account that blitzmax
is more capable, flexible and faster in itself,
it may be fairly easy to get fast results in it

jayenkai:
that's a nice trick
i usually use some dll calls to pull that off ^^
( with userlibs it's a snap really )

codersrule:
particularly the functions from the user32.dll
can make the entire blitz window:
- stretch to overlap the entire screen,
- remove the caption and borders,
- and stay on top.
so basically a 'fake fullscreen'

here's the code i use to do that:


and here's the "library" (interface) i'm using:

WindowHandling.BB (25 Dec 2010)


Mon, 17 Jan 2011, 09:36
Teasy
A few days ago I came across an old image rotation algorithm which is superfast by using 3 image skew operations.
I do not remember exactly who created the function, but I optimized it a little bit ^^

I do know that the creator was one of the brilliant minds over on BlitzCoder, which could be one of the following:
- Antony, Axeman1, BadJim, Bagels, BasicGamer2, Boomer, Chevron, CyberSeth, cR@wLy, DarkEagle, DJWoodgate, Eelco, Foppy, Gnasher, Gobble, LarsG, MutteringGoblin, ObviousDisaster, RepeatUntil, SirXriS, Skevo, World_Creator or someone else!

Anyway, here's the function with example program
I have inlined any required stuff from external files for your convenience ^^


copyrectrotaskewopti 1e2.bb

Fri, 19 Feb 2021, 06:55
spinal
looking at rotation again, trying to figure put what all of the magic numbers are in my above post

-=-=-
Check out my excellent homepage!
Fri, 19 Feb 2021, 08:04
Jayenkai
Making me wish, again, that I had quick and easy Blitz access..
Any degree, or are we talking 90 degree rotations?!

-=-=-
''Load, Next List!''
Fri, 19 Feb 2021, 08:12
spinal
I've got most of it figured I think...
Any degree,


So far I think I just need to discover the bounding box of the rotated image as to not have to pixel the whole screen!

-=-=-
Check out my excellent homepage!
Fri, 19 Feb 2021, 09:12
Jayenkai
Surely it'd be easier to loop through the image and determine it's co-ord on the screen, rather than looping the entire screen...?
Wouldn't it?

-=-=-
''Load, Next List!''
Fri, 19 Feb 2021, 09:34
spinal
but that would leave gaps. The method would be as follows I expect...

find each of the 4 corners, rotate them around the handle, find the furthest in each direction then iterate inside the resulting box.

-=-=-
Check out my excellent homepage!
Fri, 19 Feb 2021, 09:51
Jayenkai
Do it twice over the size of the image.
First time, draw a 2x2 pixel splot, then the second time, the single pixel.
That should (!!) hopefully help fill in any sub-pixels.

-=-=-
''Load, Next List!''
Fri, 19 Feb 2021, 10:41
cyangames
Or, do it as was, then interpolate the surrounding pixels for the dead pixels, unless they're also dead pixels.

-=-=-
Web / Game Dev, occasionally finishes off coding games also!