getting info from a screen shot

[MBone]

Is it possible to get infomation from a screen shot

 

---------------------

- X -

- -

- something -

- -

- like -

- this -

- -

---------------------

 

I am trying to read text off a screen shot.

 

mbone

[rberq]

Maybe using text-recognition software?

[Yehnfikm8Gq]

There are a few applications that will do it. Google "screenshot OCR" or some combo like that.

[paul]

I like SnagIt myself, whcih can also do OCR.

[Cory]

Try using controls.

..........................................

If it's text you are after and it's in a focused dialog box or something like that, odds are the text is in a control. Simply set the variable text from the control contents. Of course you're not giving much info on the source so this might not be applicable. Avoids OCR and all that. If the control is different each time check out the Capture Control command. Essentially you can define a control at run time with this.

[Yehnfikm8Gq]

I agree with Cory to an extent about OCR. From printed page I've spent more time correcting than it would take to type. In commercial situations, I've advised people to get a Temp who will type correctly for not a lot of money. Only practical for large amounts of text.

 

I'd be interested in Paul's experience with SnagIt. It may depend on what is on the screen. If it's an HQ image I guess moderate chance, LQ very marginal; digital PCTV or PCHDTV moderate chance; interlaced PCTV poor chance (we may have gone digital but cable boxes and satellite receivers may still be connected via analog video). Web pages can usually be copied directly or if not from the Page Source; jpeg scans of text documents, moderate to poor. Did I leave anything out?

 

In images and TV picture, text is likely to be at an angle or in perspective which will make OCR ineffective. Low quality jpeg images will have artifacts like edge halo and "mosquitoes" that will throw off recognition.

[paul]

I agree with Cory to an extent about OCR. From printed page I've spent more time correcting than it would take to type. In commercial situations, I've advised people to get a Temp who will type correctly for not a lot of money. Only practical for large amounts of text.

 

I'd be interested in Paul's experience with SnagIt.

Of course, you're quite right. Consider applying OCR to 100,000 words and obtaining a result with 99% accuracy. Sounds good, eh? Most of us are happy with a 99% success rate. But 1% of 100,000 is of course 1000 - that's a lot of words to correct, especially if numbers are involved.

 

I've used Snagit's OCR in very limited circumstances - no more than a page of text. And I don't get 99% success!

[Cory]

I've had a bit of experience with OCR and one of the drawbacks is that there is always an error rate. Having said that in a few tests I've done I've found that the rate of errors in scanned documents is less than the human fingers typing. IE give a human a list of numbers to type in and they will make more mistakes typographically than OCR. Also the errors by OCR tend to be obvious like mistaking a 5 for an S and in an batch of numeric data these are east to ferret out. However duplicate entries is often the answer. When they were transcribing the Ellis Island immigration records 3 different people would manually type in the values and only if all three were in perfect agreement was the data allowed to move forward. For years bookkeepers would do this with their data entry clerks and only when both agree move forward. Unfortunately if you OCR twice you're going to get the same results.

[acantor]

If the window you want to capture is a message box, try using Ctrl + C to copy the text to the clipboard. make sure that the message box has focus, or that your script transfers focus to the window.

 

Once the text is in the clipboard, it's a snap to process.

[lokhor]

I'm having difficulty getting the text from a control that displays error messages in JDEdwards Enterprise One. I know that rational robot is able to retrieve the text from the same control but using macro express I am unable to. The class of the control is ListBox and I have tried several different methods of selecting the control and attempting to copy the text in it

 

Any help would be much appreciated.

[Yehnfikm8Gq]

It may well be that MEPro cannot do it. The fact that you've tried several methods would lead one to think you know what you are doing. I mentioned in a recent post that I could pick off specific items using AutoIt whereas MEPro could only pick off the Group Box surrounding four items. What does the Capture Window Control highlight? What information can you retrieve? The whole box? Line? Presumably you are checking the "Save the Control using the Text of the Control".

 

I dug out the AutoIt External Script I used for the example I mentioned about. Very simple:

$var = ControlGetText("Graph", "Performance History", "[CLASS:Static; INSTANCE:2]")
ConsoleWrite($var)

In the ME External Script dialog I saved the Console Output to %T[2]% and used that in ME. That simple action combined the deeper probing of AutoIt with ME's ease of data handling and display. You need AutoIt installed to have the above script run. AutoIt's InfoTool provides the necessary detail on the window.

 

 

ps it would have been better to start this as a new thread

Edited March 11, 2010 by JohnS

[Eno Nedlog]

I like SnagIt myself, whcih can also do OCR.

 

I have been trying to get ScreenOCR to work with MEP I just want to capture numbers and then past them into Excel, but seem to be having problems getting MEP to do it.

 

Using ScreenOCR manually, it picks up and recognises fairly low definition graphics very accurately.

 

http://screenocr.com/