Last time we were in Norfolk i spent a lot of time shooting with the Xpan and thinking about the composing in thirds. Which lead to things such as this:
Yesterday, going through some of the digital stuff, that never got processed, there were some other obvious attempts at composing with strict thirds. At most it’s probably supposed to be a rough guide line for proportions in pictures, but like many such guides, taking it to extremes is interesting. One of the shots was this one:
which ended up looking best cropped to 16:9. The different tones / textures of grey are somewhat pleasing. It obviously fades out as it moves up…
One thing tends to lead to another, and it occurred to me that taken to the extreme, such an image could be reduced down to three bands of colour. After some consultation with Sven, who supplied me with this:
BufferedImage intermediate = new BufferedImage(1,3,BufferedImage.TYPE_3BYTE_BGR); BufferedImage scaledImage = new BufferedImage(in.getWidth(),in.getHeight(), BufferedImage.TYPE_3BYTE_BGR); double scaleX = 1.0 / in.getWidth(); double scaleY = 3.0 / in.getHeight(); AffineTransformOp op = new AffineTransformOp(AffineTransform.getScaleInstance( scaleX, scaleY), AffineTransformOp.TYPE_BICUBIC); op.filter(in, intermediate); AffineTransformOp op2 = new AffineTransformOp( AffineTransform.getScaleInstance( 1.0 / scaleX, 1.0 / scaleY), AffineTransformOp.TYPE_NEAREST_NEIGHBOR); op2.filter(intermediate, scaledImage);
The image can be reduced to the absolute minimum:
There we have it: artistic <cough> minimalism, maximal nerdism!
Update: now with
pseudo-random noise related to the standard deviation from the average within the given third. A pseudo-randon amount of the std dev is applied to each pixel. As you can see there is more ‘noise’ in the bottom third, as there is more variation in the original image.
And, using a slightly different algorithm for colour, where the standard deviation for each of red / green / blue is separately applied (randomly):