Description

Handling of the number of pixels in ModularEncode functions is strange: being applied to a sequence of stream images like in ModularFrameEncoder::ComputeTree or to an individual image like in other cases, it overestimates the number of pixels processed due to special treatment of *total_pixels == 0 case.

If initially total_pixels = 0 then we have after the function total_pixels = 2 * (number of pixels in the image) increased by this part and GatherTreeData , in any other case we have total_pixels += (number of pixels in the image) increased by GatherTreeData (modulo edge cases).

It looks like that the intention was to not allow zero in the number of pixels as it crashes LearnTree function. However, it can be done in a more straightforward way without pixel number overestimation.

In my testing the proposed change also leads to a slightly better compression, of the order of $6\times10^{-3}$ %. However, some of the tests are failing due to increase of the file size, so I mark this as draft before I can find out the cause.

Pull Request Checklist

• CLA Signed: Have you signed the Contributor License Agreement (individual or corporate, as appropriate)? Only contributions from signed contributors can be accepted.
• Authors: Have you considered adding your name to the AUTHORS file?
• Code Style: Have you ensured your code adheres to the project's coding style guidelines? You can use ./ci.sh lint for automatic code formatting.