Incorrect source subsampling for bmp

[sghpjuikit]

Describe the bug
IIOParam.setSourceSubsampling(2, 2, 0, 0) causes incorrect reading of image if sourceXSubsampling or sourceXSubsampling is not 1. Image is rendered twice, scaled and overlapped. Happens only for bmp images. I suspect a typo in x and y coordinates somewhere in the algorithm.

Version information

1. The version of the TwelveMonkeys ImageIO library in use.
   3.8.2

2. The exact output of java --version (or java -version for older Java releases).
   For example:

IBM Semeru Runtime Open Edition 17.0.3.0 (build 17.0.3+7)
Eclipse OpenJ9 VM 17.0.3.0 (build openj9-0.32.0, JRE 17 Windows 10 amd64-64-Bit Compressed References 20220422_155 (JIT enabled, AOT enabled)
OpenJ9   - 9a84ec34e
OMR      - ab24b6666
JCL      - dc07fd49b92 based on jdk-17.0.3+7)

3. Extra information about OS version, server version, standalone program or web application packaging, executable wrapper, etc. Please state exact version numbers where applicable.
   Win 10, JDK17 IBM Semeru

To Reproduce
Steps to reproduce the behavior:

1. Read bmp image with source subsampling > 1

Expected behavior
Unscaled uncropped image

Example code

val stream = ImageIO.createImageInputStream(File("path to image"))
val reader = ImageIO.getImageReaders(stream).asSequence().first() 
reader.input = stream
reader.read(0, ImageReadParam().apply { setSourceSubsampling(2, 2, 0, 0) })

On my system, the reader's class is com.twelvemonkeys.imageio.plugins.bmp.BMPImageReader

Sample file(s)
I'll upload image if necessary. It seems this concerns every bmp image.

[haraldk]

Hi,

Thanks for reporting!

I think can reproduce the issue you describe, but only for some BMPs (24 bit/pixel BMPs).

Using the test file imageio/imageio-bmp/src/test/resources/bmpsuite/g/rgb24.bmp:

I get this result:

While the expected result should be:

(The problem is that the subsampling don't take into account that the samples are wider than the data type, ie. 24 bits/pixel is 3 bytes/pixel. This is why you see 1/3rd of the image horizontally subsampled, then the next 2/3rds non-subsampled but truncated.)

[haraldk]

Closing for now, let me know if you still have issues using the latest snapshot release.

[sghpjuikit]

Hello. I didn't think you would fix this so soon, Anyway:
I couln't find the snapshot jars anywhere, so I built it on my own and used in my project. I did have some issues but in the end, the project used the snapshot version and the issue is gone. Thx.

Btw, the logic seem suspect to me

            switch (bitCount) {
                case 1:
                case 2:
                case 4:
                case 8:
                case 24:
                    byte[] rowDataByte = ((DataBufferByte) rowRaster.getDataBuffer()).getData();
                    int bitsPerSample = bitCount == 24 ? 8 : bitCount;
                    int samplesPerPixel = bitCount == 24 ? 3 : 1;

Why are you testing the bitcount when it already is confirmed to be 24?

[sghpjuikit]

Oh, my bad, I didn't realize switch with fall-through is still a thing in older Java versions. Nvm

[haraldk]

Snapshots can be found here (link on the front page). But building yourself is also good! 😀

And yes, I considered having completely separate branches for the <= 8 bit case and the 24 bit case, but it would just be more duplicated code (and scope issues) so I kept the fall-troughs as it was.

[sghpjuikit]

I have never realized until today that the badges on top of README file are clickable! That's why I could not find the snapshots.
Thx for pointing me to this.