Re: How to improve rotation quality?

Alec Bennett wrote:
> I'm wondering if anyone has any idea how to get better results from 
> PIL's rotate() function? No matter what filter I use I'm getting very 
> jagged edges after rotating an image.
> 
> Here's how I'm invoking the filters:
> 
> pic = pic.rotate(random_rotation, resample=Image.NEAREST, expand=1)   
> pic = pic.rotate(random_rotation, resample=Image.BILINEAR, expand=1)   
> pic = pic.rotate(random_rotation, resample=Image.BICUBIC, expand=1)    # 
> the best I think

Too bad ANTIALIAS doesn't seem to be available for rotate -- darn. It 
might look better.

But to some extent, you can only get so good rotating an image.

However, I took at look at your test images -- it looked like better 
smoothing was going on inside the white border than outside, which made 
me think -- when rotating, if you are going to use interpolation, how 
does it interpolate to outside the image? It doesn't. So I tried adding 
a black background to the image first, then rotating it -- much better. 
See the enclosed version of your test code, and a rotated image.

-Chris

-- 
Christopher Barker, Ph.D.
Oceanographer

Emergency Response Division
NOAA/NOS/OR&R            (206) 526-6959   voice
7600 Sand Point Way NE   (206) 526-6329   fax
Seattle, WA  98115       (206) 526-6317   main reception

[email protected]

_______________________________________________
Image-SIG maillist  -  [email protected]
http://mail.python.org/mailman/listinfo/image-sig

from PIL import Image


im1 = Image.open("test.jpg").convert('RGBA')

size = im1.size
im = Image.new("RGBA", (size[0]+20, size[1]+20), "black" )
im.paste(im1, (10,10))

# http://www.pythonware.com/library/pil/handbook/image.htm
# NEAREST (use nearest neighbour),
# BILINEAR (linear interpolation in a 2x2 environment)
# BICUBIC (cubic spline interpolation in a 4x4 environment). If omitted, or if the image has mode "1" or "P", it is set to NEAREST.

# make different versions of the image using different filters
angle = 5
im1 = im.rotate(angle, expand = 1)    
im2 = im.rotate(angle, resample=Image.NEAREST, expand = 1)
im3 = im.rotate(angle, resample=Image.BILINEAR, expand = 1)
im4 = im.rotate(angle, resample=Image.BICUBIC, expand = 1)


images = [im1, im2, im3, im4]

#save each of the images as test-1.jpg, test-2.jpg, etc.
for counter, i in enumerate(images):
    i.save("test-" + str(counter) + ".jpg", dpi=(300, 300))