Crash when reading TIFF file

midikiman commented 7 years ago

Please fill out as much known and relevant information as possible.

Version Information:

TIFF Java Version: 1.0.1
TIFF Java Source: Central Repository
IDE Name & Version: Netbeans
Maven Version: (mvn -version) [Netbeans plugin]
Java Version: (java -version) 1.8.0_131
Platform & OS: Windows 10 x64
Other Relevant Libraries:

Expected Results:

What did you expect to happen? Expected TIFF file to load without problems

Observed Results:

What happened instead? IndexOutOfBounds exception thrown while reading the sample format:
How often does this occur? Every time with this file. Have not tried others from this source.

Output:

Any logs, errors, or output messages? --- exec-maven-plugin:1.2.1:exec (default-cli) @ amazon_planet --- TIFF Image: train_0.tif

-- File Directory --

NewSubfileType (254) LONG (4 bytes) Count: 1 Values: 0

ImageWidth (256) LONG (4 bytes) Count: 1 Values: 256

ImageLength (257) LONG (4 bytes) Count: 1 Values: 256

BitsPerSample (258) SHORT (2 bytes) Count: 4 Values: [16, 16, 16, 16]

Compression (259) SHORT (2 bytes) Count: 1 Values: 1

PhotometricInterpretation (262) SHORT (2 bytes) Count: 1 Values: 2

ImageDescription (270) ASCII (1 bytes) Count: 25 Values: [{"shape": [256, 256, 4]}]

StripOffsets (273) LONG (4 bytes) Count: 1 Values: [304]

SamplesPerPixel (277) SHORT (2 bytes) Count: 1 Values: 4

RowsPerStrip (278) LONG (4 bytes) Count: 1 Values: 256

StripByteCounts (279) LONG (4 bytes) Count: 1 Values: [524288]

PlanarConfiguration (284) SHORT (2 bytes) Count: 1 Values: 1

Software (305) ASCII (1 bytes) Count: 12 Values: [tifffile.py]

DateTime (306) ASCII (1 bytes) Count: 20 Values: [2017:04:26 23:11:27]

ExtraSamples (338) SHORT (2 bytes) Count: 1 Values: [1]

SampleFormat (339) SHORT (2 bytes) Count: 1 Values: [1] Exception in thread "main" java.lang.IndexOutOfBoundsException: Index: 1, Size: 1 at java.util.ArrayList.rangeCheck(ArrayList.java:653) at java.util.ArrayList.get(ArrayList.java:429) at mil.nga.tiff.FileDirectory.getFieldTypeForSample(FileDirectory.java:1274) at mil.nga.tiff.FileDirectory.readRaster(FileDirectory.java:1149) at mil.nga.tiff.FileDirectory.readRasters(FileDirectory.java:1109) at mil.nga.tiff.FileDirectory.readRasters(FileDirectory.java:979) at mil.nga.tiff.FileDirectory.readRasters(FileDirectory.java:931) at mil.nga.tiff.FileDirectory.readRasters(FileDirectory.java:910) at mil.nga.tiff.TiffFileTester.main(TiffFileTester.java:60)

BUILD FAILURE

Steps to Reproduce:

Run TiffFileTester on the file (or just try to load it directly)
readRasters() fails.

Relevant Code:

// Code to reproduce the problem?

        File input = new File("d:/kaggle/amazon_planet/train-tif-v2/train_0.tif");
        TIFFImage tiffImage = TiffReader.readTiff(input);
        List<FileDirectory> directories = tiffImage.getFileDirectories();
        FileDirectory directory = directories.get(0);
        Rasters rasters = directory.readRasters();

Or see TiffFileTester

Test Files:

Links to any files needed for testing? train_0.zip

Additional Information:

Any additional configuration, data, or information that might help with the issue?

midikiman commented 7 years ago

On doing some investigation into the tags, it looks like some TIFF processors play fast and loose with the SampleFormat tag. There should be the same number of entries as SamplesPerPixel, but when the samples are all of the same type only one is stored. I made a simple fix to handle it for my application, not sure how robust or general it is.

bosborn commented 7 years ago

I came to the same conclusion as you. Made a fix on the develop branch here: 591433f Fix will go out on the next release.

TIFF Image: train_0.tif

-- File Directory --

NewSubfileType (254)
LONG (4 bytes)
Count: 1
Values: 0

ImageWidth (256)
LONG (4 bytes)
Count: 1
Values: 256

ImageLength (257)
LONG (4 bytes)
Count: 1
Values: 256

BitsPerSample (258)
SHORT (2 bytes)
Count: 4
Values: [16, 16, 16, 16]

Compression (259)
SHORT (2 bytes)
Count: 1
Values: 1

PhotometricInterpretation (262)
SHORT (2 bytes)
Count: 1
Values: 2

ImageDescription (270)
ASCII (1 bytes)
Count: 25
Values: [{"shape": [256, 256, 4]}]

StripOffsets (273)
LONG (4 bytes)
Count: 1
Values: [304]

SamplesPerPixel (277)
SHORT (2 bytes)
Count: 1
Values: 4

RowsPerStrip (278)
LONG (4 bytes)
Count: 1
Values: 256

StripByteCounts (279)
LONG (4 bytes)
Count: 1
Values: [524288]

PlanarConfiguration (284)
SHORT (2 bytes)
Count: 1
Values: 1

Software (305)
ASCII (1 bytes)
Count: 12
Values: [tifffile.py]

DateTime (306)
ASCII (1 bytes)
Count: 20
Values: [2017:04:26 23:11:27]

ExtraSamples (338)
SHORT (2 bytes)
Count: 1
Values: [1]

SampleFormat (339)
SHORT (2 bytes)
Count: 1
Values: [1]

-- Rasters --

Width: 256
Height: 256
Number of Pixels: 65536
Samples Per Pixel: 4
Bits Per Sample: [16, 16, 16, 16]

Pixel x = 0, y = 0: [5264, 4412, 2859, 6621]
Pixel x = 128, y = 128: [5267, 4408, 2837, 6335]
Pixel x = 255, y = 255: [5205, 4340, 2824, 6281]

gdallocationinfo train_0.tif 0 0
Report:
  Location: (0P,0L)
  Band 1:
    Value: 5264
  Band 2:
    Value: 4412
  Band 3:
    Value: 2859
  Band 4:
    Value: 6621

gdallocationinfo train_0.tif 128 128
Report:
  Location: (128P,128L)
  Band 1:
    Value: 5267
  Band 2:
    Value: 4408
  Band 3:
    Value: 2837
  Band 4:
    Value: 6335

gdallocationinfo train_0.tif 255 255
Report:
  Location: (255P,255L)
  Band 1:
    Value: 5205
  Band 2:
    Value: 4340
  Band 3:
    Value: 2824
  Band 4:
    Value: 6281

cgohlke commented 7 years ago

I'll correct this SampleFormat bug in tifffile.py. Thanks.

bosborn commented 7 years ago

Fixed in release 1.0.2 (45aa75e)

ngageoint / tiff-java