How HaftiX works and why the patterns may turn out to be different than you expected.
I will try to present the way HaftiX works and explain any doubts concerning “why the cross-stitch pattern turns out to be different from what I wanted”. All examples will be illustrated with an image I received from the application user. The problem concerned obtaining appropriate, expected quality of the pattern.
Article for: beginners.
Author: Grzegorz Zochowski
Copying text and images without author's consent prohibited. It is allowed to refer to the content by including a link or this page in a HTML frame.
Why HaftiX selects more colours than there are on the image?The beginning is simple and intuitive – the application opens an image selected by the user.
The first sight proves that it will be possible to process the image with literally several colours. Why then it isn't so?
To answer the question, it is necessary to recall some properties of human sight.
The eye cheats on us very often and we are not aware of that. It can correct the image seen to a large extent. A simple experiment – if it is dark outside and you are sitting with lights on, what colour is a blank piece of paper? White? Our sight tells us yes but it is possible to take a picture of this paper with a camera which does not correct white balance (let's not elaborate on what it is, it does not matter now) and it will turn out that the piece of paper is probably orange, if you have a bulb or a lamp emitting “warm” light on. It happens so because we have learnt that the paper is white and how white looks like in artificial light and we see no difference.
Let's go back to our image which at the first sight consists of several colours and it is obvious to us. Let's see how it looks zoomed:
Another reason of too large number of colours in a pattern results from the fact that the cross-stitch pattern consists of quite large single-coloured squares. Let's assume that the target pattern will have 10 x 13 crosses. Graphically, each cross will correspond to a field of quite large surface area:
Now choose any field in the middle of the image and try to tell what colour it should have. Difficult, isn't it? The application did the task the following way:
It is a result of ordinary mathematical calculations and although the effect is unsatisfactory to a large extent, it is impossible to do it any other way. Going back to the multitude of colours in a pattern – if there are several colours in a given field (for example, a red skirt, a grey shirt, a black hand and white background), the application calculates an average of all those colours and a new colour is created. It happens many times during the transformation of the image into a pattern and you need to take into account the fact that intermediate colours between those present on the image will appear.
The image discussed does not contain another phenomena which are frequent in case of pictures which later cause astonishment with their result. Typically, it is about unnatural colours of faces. Not many people can notice that this is the way they are on the photo, since cameras of low class and amateur photographers are not able to take a photo in such a way that the colour of skin is warm and not livid or even purplish. Therefore, if you want to have a nice pattern, you should first graphically process the photo. It is a separate and quite complicated issue.
In what way HaftiX minimizes the amount of colours?A problem separate to obtaining too large number of colours is... their too small number. If we want to cross-stitch a rainbow, we will surely not find threads in colours which would be necessary. So, the question arises – how to transform a many-hued images into several or several dozen colours to get the best effects?
There is no universal answer to this question. The application is equipped with some compromising solutions which in some cases give worse results. We can consider a situation that you need, for example, 20 thousand various colours. It is not a large amount, a photo rarely contains less. To replace the amount with several dozen colours, we can do the following: group all hues into groups containing colours similar to each other (for example, shades of light green, dark grey, etc.) and from these groups choose one that will replace all others. This method may seem good but only for short time. In case of 10 colours you need to create 10 groups containing 2 thousand colours on average. It is easy to imagine how much information you will lose if you replace 2 thousand colours with one.
Another method consists in finding colours which are the least frequent and replacing them with the most frequent ones. The procedure in our case would consists in finding 10 most frequent colours and replacing all of them with one of them.
The abovementioned methods have their significant disadvantages. The first one selects replacement colours for the most frequent shades in a worse way. The other method – selects colours for the least frequents ones in a worse way. The HaftiX application uses the first method since the other was not the best for processing photos of faces (the proper colour of lips, eyes was lost and these are small areas but of characteristic colour). Improvements of this mechanism are in progress and in the future it should be more "smart".
What can you do?To get a desired effect, you may transform the image using a larger number of colours first, for example, 16. You will get the following pattern:
Using the colour replacement tools, change some colours present in the jacket and skirt into dominating colours. Please pay attention to the man's doublet. The darker colour has been replaced with lighter one. If you want to do similar operation, I recommend to learn and train how to use the colour replacement tools. I strongly advise against using the “fill in” tool which may leave single crosses somewhere.
The result is as follows:
When using the application skillfully, it is possible to prepare such pattern within 10 minutes. I encourage you to read the manual, try new tools, practice work with patterns. It will surely result in better patterns.