Is it possible to search files by text shown in images?

Help & Support
1

I was wondering that is there a possibility to search images by text that is shown in the images? I need to search a hard drive full of files, for some files (text files or images) that have certain content text. No matter that are those files text files, documents or images (screenshots etc).

2

Not in general, although it could be done with an OCR IFilter. Maybe one already exists, since IFilters are also supported by parts of Windows and Office.

3

So.. the answer is no. Haven't found any tools for this with which I could index my NAS and search my gazillion screenshot files with search text that is contained in them. Phew, I'm in trouble..

4

Tesseract and Opus let you quickly build a text mirror of your image collection which can easily be searched with Opus or any convenient search tool.

Resources:

A button for Tesseract can be as simple as

@nodeselect
@nofilenamequoting
"/programfiles\Tesseract-OCR\tesseract.exe" "{filepath}" "{filepath}.txt"
5

Thanks for the reply. If I understand correctly, that OCRs one selected file to text when a button is pressed? How to extract those .txt files from the whole NAS to a certain folder and just search that folder then?

I asked AI about that and it suggested:

# Define paths
$srcFolder = '\\NAS\MyPictures\' 
$dstFolder = 'C:\OCR-Results\'

# Get all jpg, jpeg and png files in source directory and subdirectories
$files = Get-ChildItem -Path $srcFolder -Recurse -Include *.jpg, *.jpeg, *.png

# Loop through all files
foreach ($file in $files) {
    # Define output file name
    $outputFile = Join-Path $dstFolder ($file.BaseName + ".txt")

    # Run Tesseract OCR and output result to defined file
    Start-Process -FilePath "/programfiles\Tesseract-OCR\tesseract.exe" -ArgumentList "`"$($file.FullName)`" `"$outputFile`""
}
6

Same principle in Opus. Get a list of files via Find or Flatview/Filter and run Tesseract on it. You can use codes in the button to rebuild the folder tree.

@nodeselect
@nofilenamequoting
"/programfiles\Tesseract-OCR\tesseract.exe" "{filepath}" "C:\OCR-Results\{filepath|noroot}.txt"

Nothing wrong with Powershell, of course. You will probably need to use the Windows notation for the path to tesseract.exe.

7

I ended up with this, does this look usable and correct to you:

@nodeselect 
@nofilenamequoting 
@runonce:@set dirname={filepath|root}\OCR-files
@runonce:@set relpath={filepath|..}
@runonce:@set relpath={$relpath|noterm}

CreateFolder NAME "{$dirname}{$relpath}" READAUTO=no 

"path\to\tesseract.exe" "{filepath}" "{$dirname}{$relpath}\{file|noext}.txt"
8

Try

@nodeselect 
@nofilenamequoting 
@runonce:@set dirname={filepath|\}OCR-files
@runonce:@set relpath={filepath|..|noroot|noterm}

CreateFolder NAME="{$dirname}\{$relpath}" READAUTO=no 

"path\to\tesseract.exe" "{filepath}" "{$dirname}\{$relpath}\{file|noext}.txt"
9

Using this now:

@nodeselect 
@nofilenamequoting 
@runonce:@set dirname={filepath|\}OCR-files
@runonce:@set relpath={filepath|..|noroot|noterm}

CreateFolder NAME="{$dirname}\{$relpath}" READAUTO=no 

"C:\Program Files\Tesseract-OCR\tesseract.exe" "{filepath}" "{$dirname}\{$relpath}\{file}"

I noticed that the OCR makes quite a lot of mistakes, which is a bit surprising. When there is clear text that says for example "Liike" it outputs "Like" and on some words it is missing multiple characters. I also noticed that it won't output umlauts. I guess it requires configuration somewhere. Have two kids to burn my time, so don't have the time to dig deeper. :sweat_smile:

10

Adding languages to the command line should help, at least when handling special characters. It's been a while since I last used Tesseract, so my knowledge is not current.

11

awesome! thanks for this guys!

yeah it makes a lot of mistakes... but the 3 other softwares ive been using do a well. each one will come up with different misspellings or random symbols. haha

its nice having this automatic so i dont have to spend time opening other software!