Improved Diplopia Solution

[woodjohndavid]

I previously created a pull request from my branch JDWDIPLOPIA. This solution, while better than without it as far as diplopia is concerned, was a limited solution.

I am now creating a pull request from my branch JDWDIPLOPIA2. I believe that this is a more complete solution to the diplopia issue.

[stweil]

I don't see a commit which changes the Tesseract code. Is something missing?

[woodjohndavid]

OK thanks @stweil

This may leave you somewhat concerned about my code changes, but I can assure you that I am a competent developer. I just never use GitHub other than with Tesseract, hence incompetent in that regard.

Note by the way, similar to the other pull request I generated, there are some new configuration values that can only be set in code as it stands, but should be made into available settings. I have not yet figured out the mechanism for doing that. If the diplopia changes I am proposing turn out to be useful, hopefully someone else familiar with the settings approach could take care of that. These configuration values are:

• bool kRemoveDiplopia - if true, enables diplopia removal functionality. If false, my changes have no effect
• int kMaxDiplopiaGap - number of timesteps apart to be considered diplopia, default 2

[amitdo]

Apart from testing that this patch has a positive effect on the diplopia issue, people should test if there is no negative effect in other places, like dropping of correct characters.

[woodjohndavid]

Please note that this change is likely not appropriate for those using Tesseract for natural language recognition using relevant dawgs. It is primarily intended for those (like myself) using Tesseract to scan technical data, looking for exact character by character recognition.