DICOM RLE Codec written in Rust

Full implementation of a DICOM RLE CODEC that runs in WebAssembly.

See the dicomrle-wasm repository for the WebAssembly version

Actively being developed (Jun 3, 2020)

• Implement Decoder
• Implement Encoder
• Create Performance Bechmarks

• DICOM Standard on RLE Encoding
• CornerstoneJS JavaScript Source
• GDCM C++ Source
• DCMTK C++ Source
• ClearCanvas C# Source

The decode() api requires that the caller presize the decode Vector to be the expected size of the decoded image. This was done so the decoder could properly detect incomplete decodes. Since this decoder is designed to decode DICOM RLE Images, the caller should have access to the DICOM Header which contains the attributes needed to calculate the size of the decoded image buffer (specifically - rows, columns, bits allocated and samples per pixel). Presizing the decode Vector also improves the performance as the vectors capacity never has to be expanded (which would involve a copy operation)

RLE decoding cannot be streamed on both the input and output simultaneously without fully buffering either the output or input because RLE encoded images typically contain multiple segments which must be interleaved. It is theoretically possible to stream either the input or output, but not both simultaenously.
Streaming the output would be the better choice since it is bigger, but this would complicate the decoder logic as it would have to decode all segments concurrently so they could be interleaved. I have decided not to implement streaming support as it would make things more complicated and I currently have no need for streaming functionality.

This library makes minimal use of the standard library. Use of the standard library could probably be eliminated which would produce slightly smaller WebAssembly binaries. The public interface does not depend on the standard library so accomplishing this is just a matter of modifying the internal implementation.

This library does not utilize unsafe code except for the the decode_u16() and decode_i16() functions which use it to cast the decoded buffer from u8 to u16 or i16 to avoid an additional allocation and copy that would otherwise be required. If you want to avoid unsafe code and need to decode i16 or u16 images, you can still call decode() which will work properly even if the decoded data is i16 or u16, but you will have to convert it yourself.