I find Scantailor a very useful application, and more user-friendly than ScanKromsator. Still, it is always possible to improve. A few suggestions of possible improvements:
- When you open the "about" box, there is no indication how to close it. I know, typing "Escape" closes the box, but people like to see an icon or button which says "click me to close the about box".
- Perhaps it would be useful to add a "Help" in the menu, even if clicking on "Help" just goes to the user manual on the sourceforge wiki?
- Sometimes the preview window shows only a few pages. To see all pages, I have to change the order. (e.g. If I am in "natural order", by clicking on "order by increasing width", and then go back to "natural order"). Is this something I am doing wrong?
- I can go to the next/previous page by typing "q" or "w", but I do not see this documented anywhere.
What do you think?
ST suggestions and feature requests go here.
Moderator: peterZ
-
- Posts: 4
- Joined: 04 Mar 2014, 00:53
Re: ST suggestions and feature requests go here.
Nice to know. Thank you for pointing that out.plecosaurus wrote:go to the next/previous page by typing "q" or "w"
Anyone know of other undocumented keyboard shortcuts?
Re: ST suggestions and feature requests go here.
There is a tip on the status bar when you mouse over the list of thumbnails.plecosaurus wrote:I can go to the next/previous page by typing "q" or "w", but I do not see this documented anywhere.
Scan Tailor experimental doesn't output 96 DPI images. It's just what your software shows when DPI information is missing. Usually what you get is input DPI times the resolution enhancement factor.
Re: ST suggestions and feature requests go here.
Hello, everybody,
first of all I want to say thank you to Tulon for this fantastic piece of software called ScanTailor. Tulon, I'm a big fan of yours, and if you ever come up to Stoke, all drinks are on me, and you can sleep with my sister as well. (Only joking, ha ha; I haven't got a sister.)
I've been having a look at the code in the hope that I could co-operate in this project, but unfortunately my level of C++ is not good enough. I've also been testing the new experimental automatic dewarping tool, and I have a couple of suggestions to make. I may be talking total rubbish, but who knows.
First of all, it seems the tool is trying to dewarp and dekeystone at the same time. Wouldn't it be easier, for example, to create a tool for images that are only keystoned, and if this succeeds then go for the dewarping? Wouldn't it make sense to have two separate tools from the computational point of view as well? I'm sure dewarping is a lot more computationally-intensive than dekeystoning and sometimes only dekeystoning is needed.
My second suggestion is, it seems the program is not very good at detecting the direction of lines. If I remember well, Tulon was trying to fit lines by quadratic spline interpolation, which resulted in a wavy line going up and down with exponents, subscripts, etc. Here is my suggestion for images that are only keystoned:
1- Detect all characters on a line
2- Select lowest pixel of each character
3- Fit a regression line by least squares.
4- Discard outliers--this would get rid of g's, p's, y's, apostrophes, exponents, subscripts, etc.
5- Fit a regression to the remaining pixels. This would be our baseline.
6- Make all baselines horizontal.
7- Move all non-outliers to sit right on top of baseline where they originated. Keep outliers glued to nearest non-outlier.
The calculation of two regression lines would also require a lot less computation than the interpolation of ten or fifteen second-degree polynomials. Of course, images, mathematical formulas, etc, could mess the whole thing up, so they would need to be detected and dealt with appropriately.
If this is a lot of rubbish, I apologise for wasting your time, Tulon; I'm sure you're a very busy person. Thanks for all your work once again.
Jack
first of all I want to say thank you to Tulon for this fantastic piece of software called ScanTailor. Tulon, I'm a big fan of yours, and if you ever come up to Stoke, all drinks are on me, and you can sleep with my sister as well. (Only joking, ha ha; I haven't got a sister.)
I've been having a look at the code in the hope that I could co-operate in this project, but unfortunately my level of C++ is not good enough. I've also been testing the new experimental automatic dewarping tool, and I have a couple of suggestions to make. I may be talking total rubbish, but who knows.
First of all, it seems the tool is trying to dewarp and dekeystone at the same time. Wouldn't it be easier, for example, to create a tool for images that are only keystoned, and if this succeeds then go for the dewarping? Wouldn't it make sense to have two separate tools from the computational point of view as well? I'm sure dewarping is a lot more computationally-intensive than dekeystoning and sometimes only dekeystoning is needed.
My second suggestion is, it seems the program is not very good at detecting the direction of lines. If I remember well, Tulon was trying to fit lines by quadratic spline interpolation, which resulted in a wavy line going up and down with exponents, subscripts, etc. Here is my suggestion for images that are only keystoned:
1- Detect all characters on a line
2- Select lowest pixel of each character
3- Fit a regression line by least squares.
4- Discard outliers--this would get rid of g's, p's, y's, apostrophes, exponents, subscripts, etc.
5- Fit a regression to the remaining pixels. This would be our baseline.
6- Make all baselines horizontal.
7- Move all non-outliers to sit right on top of baseline where they originated. Keep outliers glued to nearest non-outlier.
The calculation of two regression lines would also require a lot less computation than the interpolation of ten or fifteen second-degree polynomials. Of course, images, mathematical formulas, etc, could mess the whole thing up, so they would need to be detected and dealt with appropriately.
If this is a lot of rubbish, I apologise for wasting your time, Tulon; I'm sure you're a very busy person. Thanks for all your work once again.
Jack
Re: ST suggestions and feature requests go here.
jacko,
I think your suggestions are premature at this point.
First of all, I still have quite a few improvements for text line tracing in mind.
Second, I don't think we would gain much performance-wise. I haven't profiled the dewarping code yet, but I doubt snake evolution takes much time.
Anyway, if it turns out to be worth it, straight line tracing is not hard to implement.
I think your suggestions are premature at this point.
First of all, I still have quite a few improvements for text line tracing in mind.
Second, I don't think we would gain much performance-wise. I haven't profiled the dewarping code yet, but I doubt snake evolution takes much time.
Anyway, if it turns out to be worth it, straight line tracing is not hard to implement.
Scan Tailor experimental doesn't output 96 DPI images. It's just what your software shows when DPI information is missing. Usually what you get is input DPI times the resolution enhancement factor.
Re: ST suggestions and feature requests go here.
Ha ha, so I am talking rubbish... Thanks for being polite enough not to say it directly. Looking forward to your next versions of ScanTailor.
Cheers,
Jack
Cheers,
Jack
-
- Posts: 4
- Joined: 04 Mar 2014, 00:53
Re: ST suggestions and feature requests go here.
I've looked more closely at the "About" box, and there's a difference between the mac and windows versions of the program. The windows version shows an "About" box with icons to close the window (The small icon in the upper right corner with a cross). The mac version doesn't show the icons to close the window. Strange.plecosaurus wrote:- When you open the "about" box, there is no indication how to close it. I know, typing "Escape" closes the box, but people like to see an icon or button which says "click me to close the about box".
Re: ST suggestions and feature requests go here.
I'll try adding a quick workaround via Qt Designer. Would a close button on the bottom work?
- daniel_reetz
- Posts: 2812
- Joined: 03 Jun 2009, 13:56
- E-book readers owned: Used to have a PRS-500
- Number of books owned: 600
- Country: United States
- Contact:
Re: ST suggestions and feature requests go here.
Hey ST hackers, I thought you might like to see this:
http://diybookscanner.org/forum/viewtop ... 8560#p8560
http://diybookscanner.org/forum/viewtop ... 8560#p8560
Re: ST suggestions and feature requests go here.
The mailing list is pretty slow, I must say. But anyways, here's a little something for those Mac people:
No idea on how to commit it, though...