Olympus i-SPEED 2, Instructions Manual

Page: 95 / 116

Understanding ispeed
93
i
--SPEED LT,
i
--SPEED 2
`
JPEG Compression
Although the Olympus
i
-- SPEED LT,
i
-- SPEED 2 contains a
state--of--the--art microprocessor system, there are still limitations
on the speed with which data can be written to a Compact flash
card or an Ethernet channel. Even the most modern memory cards
have a limited capacity, and the amount of video that can be saved
in any one test session is often restricted.
In order to overcome these two issues, the Olympus
i
-- SPEED LT,
i
-- SPEED 2 contains a piece of electronics called a JPEG
Compressor. This device can take live frames of video and
compress them from their raw “bitmap” format into a more tightly
compacted format called JPEG (named after the Joint Picture
Expert Group who invented the format and techniques involved)
The term “bitmap” means that for every pixel in the image, three
numbers are stored in the file, one each for red, green and blue. In
this way, every pixel is fully described and the storage of the image
is completely accurate to the original. On the other hand, this is
extremely wasteful. If this page were to be stored as a bitmap, the
file would contain a large amount of duplicated information,
because there is a large amount of pure flat white to be stored.
JPEG compression involves breaking the image into 8 x 8 or 16 x
16 pixel blocks and examining each block in turn to see if there is
any redundant information. Once this has been established, a
decision can be made about exactly what constitutes redundancy,
and so the compression ratio may be varied. (For the more
technical reader, this process is based on Discrete Cosine
Transformation followed by Huffman Encoding.)
Naturally, to achieve a higher compression ratio, some information
may be omitted, and although only the least relevant information is
omitted, some loss of image data does occur. This is normally
seen as faint speckles around highly detailed sharp edges in the
image. The higher the compression rate, the more loss occurs. As
a guideline, 7:1 compression produces images which are basically
indistinguishable from the original. 15:1 (the most common ratio for
digital cameras) produces losses which are only visible by
someone who is deliberately looking for artefacts under high zoom.
65:1 produces artefacts which are occasionally visible, but do not
detract from the appearance of the image.