Regression: gif not scraping in 5.2 that was OK in 5.1

[jacekkopecky]

Environment

• Tesseract Version: v5.2.0.20220712
• Platform: Windows 10, x64

Current Behavior:

tesseract.exe v 5.2 only scrapes top line (window title) from this gif image:

Expected Behavior:

It should scrape all text.

Tesseract v5.1.0.20220510 worked as expected, and both 5.1 and 5.2 work as expected with this equivalent png image:

Suggested Fix:

It might have something to do with a different DPI estimation: 5.2 estimates resolution 132, while 5.1 estimated 168. However, running 5.2 with --dpi 168 does not seem to fix anything.

[amitdo]

What's the output of tesseract -v for each version?

I want to see if the leptonica and gif libraries being used are exactly the same.

[jacekkopecky]

tesseract v5.1.0.20220510
 leptonica-1.78.0
  libgif 5.1.4 : libjpeg 8d (libjpeg-turbo 1.5.3) : libpng 1.6.34 : libtiff 4.0.9 : zlib 1.2.11 : libwebp 0.6.1 : libopenjp2 2.3.0
 Found AVX512BW
 Found AVX512F
 Found AVX2
 Found AVX
 Found FMA
 Found SSE4.1
 Found libarchive 3.5.0 zlib/1.2.11 liblzma/5.2.3 bz2lib/1.0.6 liblz4/1.7.5 libzstd/1.4.5
 Found libcurl/7.77.0-DEV Schannel zlib/1.2.11 zstd/1.4.5 libidn2/2.0.4 nghttp2/1.31.0

tesseract v5.2.0.20220712
 leptonica-1.78.0
  libgif 5.1.4 : libjpeg 8d (libjpeg-turbo 1.5.3) : libpng 1.6.34 : libtiff 4.0.9 : zlib 1.2.11 : libwebp 0.6.1 : libopenjp2 2.3.0
 Found AVX512BW
 Found AVX512F
 Found AVX2
 Found AVX
 Found FMA
 Found SSE4.1
 Found libarchive 3.5.0 zlib/1.2.11 liblzma/5.2.3 bz2lib/1.0.6 liblz4/1.7.5 libzstd/1.4.5
 Found libcurl/7.77.0-DEV Schannel zlib/1.2.11 zstd/1.4.5 libidn2/2.0.4 nghttp2/1.31.0

[zdenop]

It seems like there is a problem with thresholding/binarization :

tesseract i3940.gif i3940.gif get.images produced:

tesseract i3940.png i3940.png get.images produced:

By the way: gif has 256 colors (8 BitsPerPixel), and png version has 16,7 Million colors (32 BitsPerPixel)

[amitdo]

18fb5aa

[zdenop]

@amitdo : that commit is not the problem: the input gif image has already an 8-bit depth, so it just makes a copy of it.

[amitdo]

5.1.0...5.2.0

[zdenop]

ok. I got it: problem is causing Colormap.

[zdenop]

Explanation of problem/dilemma:

1. removing a colormap from a 8bit color image returns 32bit image => output pdf file size will increase - see issue PNG images being reencoded unnecessarily in PDFs #3092.
2. not removing colormap cause problem on images like above (interesting is that also image from issue 3092 is the 8bit with colormap)