As I’m only using 7 colours each TinyPic is easily represented by an indexed GIF image.  Normally, GIF images have an index of 256 colours, JPEGs up to 16million colours.  The index of TinyPic GIFs, however, only needs to have 7 colours.  So what I’ve done is forced my own 7 colour index onto whatever image is presented to my system.  That way, PHP does the colour matching for me, and I didn’t have to worry abo9ut matching up RGB values.

The way I have forced this seven colour index onto images was the trickiest part.  I imagined that PHP would have functions to allow me to build up my own index from any RGB values I wished.  That wasn’t the case unfortunately.  To get around this I’ve used the PHP function imagecopypalette() which lets you force another images index onto an image.  So I created up a seven colour gif (here it is here, look closely- ) and copied the palette of that image onto uploaded TinyPics.  Strangely, this palette copying function doesn’t seem to work on JPEG images, so for those to work I’ve had to convert from JPEG to GIF on the fly.

I’ve done a little testing by manually changing colours and it works on tyat level.  Next, I need some real testing using photographed Tinypics.

As you may have read in my previous post, I think I’ll need to make the grid 3×4 (or vice versa) in order to store enough information.  That gives me 12 colour squares, or six pairs.  So the simplest way to do this is to use colour pairings to represent characters i.e. a red square paired with a green square represents ‘a’, a blue square followed by a magenta square represents ‘b’.  It may not be all that clever, but using the six colours shown in previous posts gives me 36 combinations which is exactly what I need.

I need to have a representation for ‘null’ however, as tinyurl.com/6 does not equal tinyurl.com/000006. I can round this by introducing white and use pairings with white to represent nothing.

In my previous post you can see the investigation I’ve done into colours and how they are represented when printed and photographed.  What this research has shown is that I will have to use RGB values of either 0 or 255.  If I try to use numbers in between then I could not guarantee that my system will be able to tell them apart.  Essentially this means that my RGB values are now in binary, a colour’s red, green or blue value will have to be either ‘on’ or ‘off’.  Each colour can be represented by a three bit string (blue would be 001 for example as it has no red, no green, but does have blue).

This means that a grid of nine colours gives me a 27 bit string.  This gives 134,217,728 (2^27) combinations.  If I’m to use tinyURL.com’s API, and in turn use six alphanumeric characters then I need 2,176,782,336 (36^6) combinations.  Unless I have a ‘eureka’ moment I’m looking at adding another row of colours.

The figures below have been taken from images taken by the camera on a Nokia N95, a five megapixel camera which I’d consider to be one of the higher quality cameras found on mobile phones today.  The colours researched and their respective RGB values are below.

Once printed and photographed, the RGB values of the colours are shown below.

In good light with no flash-

And in bad light with a flash as shown below-

As you can see from the graph above, colours photographed in bad light don’t keep their RGB values very well, and the above values are with the “good” camera.  This means I will have to use colours at the extremes of the RGB scales and nothing in between i.e. Magenta is supposed to have red and blue values of 255, but once photographed these colours drop to less than 100, if I tried to use red and blue values of, say, 127 then they would drop so much it would become on a par with the green value (which has grown from 0 to 45).