Making Do With Less Data

[Winston]

Lab1 wrote:

On 6/25/2011 11:15 AM, Bret Cahill wrote:
If you wanted to make an image you could simply omit or average every
other pixel.

Is there a short list on the most popular techniques for making do
with less data?

Here's an idea...
First figure out how many different colors the image has.
Store just one pixel of each color in the file (huge savings right there!)
Then map the position of each color in x, y coordinates.

When loaded the mostly text file simply copies a color pixel and puts it
in all the locations specified in the coordinates.

Termed 'raster to vector conversion'.

http://en.wikipedia.org/wiki/Raster_to_vector

Generally results in a > 10 x file *expansion* compared to jpg.

--Winston

 

[John Larkin]

On Wed, 29 Jun 2011 10:56:37 -0400, Lab1 <.@...> wrote:

On 6/25/2011 11:15 AM, Bret Cahill wrote:
If you wanted to make an image you could simply omit or average every
other pixel.

Is there a short list on the most popular techniques for making do
with less data?

Here's an idea...
First figure out how many different colors the image has.
Store just one pixel of each color in the file (huge savings right there!)
Then map the position of each color in x, y coordinates.

When loaded the mostly text file simply copies a color pixel and puts it
in all the locations specified in the coordinates.

That would work, but would quantize the colors. I'd guess that 8 bits
per pixel, doing color and brightness from a dictionary, would look
sort of OK, and that would save some space.

The Sherlock Holmes mysteries, on PBS lately, sure look like the
colors and brightnesses are coarsely quantized.

JPEG compresses between 10:1 (looks good) and 50:1 (doesn't) so must
be much more clever.

John

 

[Jasen Betts]

On 2011-06-29, Winston <Winston@BigBrother.net> wrote:

Lab1 wrote:
On 6/25/2011 11:15 AM, Bret Cahill wrote:
If you wanted to make an image you could simply omit or average every
other pixel.

Is there a short list on the most popular techniques for making do
with less data?

Here's an idea...
First figure out how many different colors the image has.
Store just one pixel of each color in the file (huge savings right there!)
Then map the position of each color in x, y coordinates.

When loaded the mostly text file simply copies a color pixel and puts it
in all the locations specified in the coordinates.

Termed 'raster to vector conversion'.

http://en.wikipedia.org/wiki/Raster_to_vector

Generally results in a > 10 x file *expansion* compared to jpg.

it depends on the image, it works best for "line art" where there are
large geometrically related blocks of colour, I've seen a 30x reduction,
it could probably have been tweaked to give even more.

I converted a 1.2M JPEG logo to EPS using basically the default
settings in inkscape resulting in pdfs being 30 times smaller
loading faster and looking better.

--
⚂⚃ 100% natural