0 Header 2 bytes char AC
2 Major version 1 bytes Byte
3 Minor Version 1 byte Byte
4 Images Count 2 bytes Integer Nº of images contained
6 Unknown int 2 bytes Integer
{
Size X 2 bytes Integer Size of field
Size Y 2 bytes Integer Size of field
Frame Len 2 bytes Integer Length of field data
Frame Data Frame Len bytes Byte Image Data compresed in RLE version
}
- Palette 1024 bytes RGB 256 color palette

Atleast that’s what I’ve found when I was trying to make a parser for SPR file.

{ Image (Header offset +14)
10 - Size X - 2 bytes Integer - Size of field (LB HB format)
12 - Size Y - 2 bytes Integer - Size of field (LB HB format)
14 - Len Data - 2 bytes Integer – Length of field data (LB HB format)
16-(16+Len Data) - Frame data*
}
*Frame data are actual image pixels with values referencing the indexed colors in the palette except when they have a value of “00” (the background color index). When background pixels are encountered the next pixel will either be another background pixel if “00” or the count of the number of time to repeat the display of the background pixel if not “00”. This is done to achieve some level of compression. Example:

A0 00 05 B0 = A0 00 00 00 00 00 B0 (00 x 5)
A0 00 00 B0 = A0 00 00 B0

Compression runs left to right and can span more than 1 row, so do not be surprised to see background pixels drawn at the end of one row extending beyond the end of the row and wrapping to the next. Only background pixels can be compressed.

Also, the first color in the palette is always the transparency color in case you did not know that. I was going to post this on the eAthena forums, but I figure it makes more sense here since you are the file format repository.