Citation, DOI, disclosures and article data
Image compression is a technique applied to digital images to decrease the amount of space required to store an image and increase the speed with which the image can be retrieved or transmitted.
Compression may be lossless or lossy:
lossless compression retains all original data, for example PNG and TIFF formats can be saved with lossless compression
lossy compression, in contrast, is defined by the permanent loss of data during compression, which typically results in smaller file sizes; JPG is a typical example of a lossy image format
Compression is possible due to coding redundancy, image redundancy and observer limitation:
refers to unnecessary inefficiency in the coding system used for data
for example, if 1 byte (8 bits) is used to represent the level of grey in each pixel for an image that only contains 16 levels of grey, then 4 bits will be redundant; a pixel value of 15 could be expressed using 4 bits as ‘1111’ instead of using 8 bits as ‘00001111’
refers to the suboptimal encoding of pixel data, particularly when there are repeated patterns of pixels, for instance, areas of uniform color; a line of 8 pixels that all have a grey value of 9 would normally be expressed as ‘99999999.
this redundancy could be compressed using a control character (*) indicating the subsequent two digits represented A) the number of repeats followed by B) the value being repeated (i.e. *89)
run-length encoding uses this principle to compress data
refers to instances where the observer is unable to perceive beyond a certain level of detail, rendering additional detail beyond this point redundant
for example, the human eye can only perceive a maximum of approximately 900 shades of grey (from black to white) 1
the 12-bit encoding of many DICOM images (212 = 4096) is therefore in some circumstances redundant; note that windowing and levelling allow only parts of this range to be shown; similarly spatial resolution may be redundant; a 16 megapixel image shown on a 4 megapixel display is redundant (unless zooming and panning which may take advantage of this extra data)
A variety of techniques can be used to perform image compression 2 and compression of medical imaging is subject to international standards 3.
The level of compression achieved can be quantitatively expressed as the compression ratio (N1:N2) where N1 denotes the file size of the uncompressed data and N2 denotes the file size of the compressed data. A higher compression ratio will therefore produce a smaller compressed dataset.