Question

I am just processing an image, looking at things such as colour and contrast. How ever my issue is analysing the complementary colours in the image and trying to do this efficiently.

Firstly I have got the pixel rgb. I have then converted to HSV and increase the hue and return to rgb, hence obtaining the complementary colour. I am then looking at its closet neighbours to see if any of these are complementary colours. How ever given these are pixels, it is rare one would find the central pixel to be a complementary – hence I do not feel this is very efficient.

Or another idea… to segment the image in accordance to colour regions and work out the distance from one region to another if there is a region with the complementary colours.

Any ideas and any ideas on how to efficiently code this?

Thanks

Answer 1

So I eventually worked out roughly how to do this, either by a very slow way or a slightly faster way:

1. segment image into colour regions
2. calculate complementary of colour region by adding 0.5 to the hue of the colour
3. look at closest neighbours to complementary aswell, as we may not have exact complementary colour present
4. calculate euclidean distance from the segment to the complementary segment (if this exits) and calculate 1/ED – this will be 1 if close and nearer to zero if far away, so acts like a weight.
5. calculate proportions in segment to weight pairing of complementary colours accordingly.

Otherwise

1. As opposed to segments do this for each pixel to every other pixel