07-10-2005, 08:33 PM
I had a 189Kb long BMP picture in 24bit format with a total of 128 colors (I knew it that it has that number of colors since i reduced the color depth, but then i saved it as 24 bit). After i fed that BMP with one of my programs converting to a PCX style indexed format capable to hold maximally 256 colors. It worked, the image correctly saved (of course without any data - loss since it had 128 colors as i said above; it was 18Kb). Then for some reason i tried to feed the two images with a ZIP and a RAR archiver - possibly because i wanted to know how effective is that PCX algorithm. The result with ZIP:
The original BMP compressed to 14061 byte while my format compressed to 10996 byte.
Of course this difference can not be because of the headers: the 24bit BMP format has only 54byte of header data. Strange, not?? If my format was ended up being larger after the compression than the original in compressed, i would have not wondered about anything: i just put some extra mess in it. But this is the opposite effect: my format could be compressed better than the original! This means that the compression program misses a lot of possible optimizations during compressing! (It's size was 128% of what i got with ZIPping after the lossless conversion)
With RAR i got the same effect: my format "gave a hint" to the compression algorithm! Is not it strange? If it was just a few byte, i would not really pay attention to it, but 128% is way too much to simply ignore. I had never experienced such thing before although i tried a few things with compression softwares. For example converting to PNG or GIF and compressing in various ways, but never got that much difference.
At the side of my format it was good to experience such a thing: it's size was just 161% of the compressed form, not to mention that it can be used up well in real - time as the decompression can be as fast as plotting pixels from a simple bit - map, so it can save up huge amounts of memory (i first used this method with "Kicsi Kocsi" - Little Car 1.50 at about three years ago. It used PCX style compression in real time).
After all i think PCX is a great thing. Not because of it's compression level - it is good because it can be stored in the memory so it never needs to be extracted (of course as long as it is not needed to edit it). I am working on improving this format
The original BMP compressed to 14061 byte while my format compressed to 10996 byte.
Of course this difference can not be because of the headers: the 24bit BMP format has only 54byte of header data. Strange, not?? If my format was ended up being larger after the compression than the original in compressed, i would have not wondered about anything: i just put some extra mess in it. But this is the opposite effect: my format could be compressed better than the original! This means that the compression program misses a lot of possible optimizations during compressing! (It's size was 128% of what i got with ZIPping after the lossless conversion)
With RAR i got the same effect: my format "gave a hint" to the compression algorithm! Is not it strange? If it was just a few byte, i would not really pay attention to it, but 128% is way too much to simply ignore. I had never experienced such thing before although i tried a few things with compression softwares. For example converting to PNG or GIF and compressing in various ways, but never got that much difference.
At the side of my format it was good to experience such a thing: it's size was just 161% of the compressed form, not to mention that it can be used up well in real - time as the decompression can be as fast as plotting pixels from a simple bit - map, so it can save up huge amounts of memory (i first used this method with "Kicsi Kocsi" - Little Car 1.50 at about three years ago. It used PCX style compression in real time).
After all i think PCX is a great thing. Not because of it's compression level - it is good because it can be stored in the memory so it never needs to be extracted (of course as long as it is not needed to edit it). I am working on improving this format