Dynamic Rounding

Share This
Categories: Digital Basics, Images

Dynamic Rounding is a mathematical technique for truncating the binary word length (the number of 1s and 0s) of pixels to a specified number of bits. Rather than simply ignoring the lower bits, it uses their information to control, via a randomizer, the dither of the LSB of the truncated result. This effectively removes the artifacts, such as banding, that could otherwise be visible. Dynamic Rounding is non-cumulative on any number of passes and produces statistically correct results. Earlier attempts at a solution have involved increasing the number of bits (e.g. from 8 bits to 10 bits), making the size of LSBs smaller but not removing the inherent problem.

Some form of limiting the number of bits is required as there are many instances in digital systems where a number, representing a pixel value, uses more bits than the system normally accommodates. For example, a nominally 12-bit system might have a problem handling a 24-bit word. This has to be rectified in a way that will keep as much information as possible and not cause noticeable defects even after many processes. A common example arises in image processing which often involves multiplying the values of co-located pixels in two different images, as in digital mixing. Assuming the equipment is nominally 12-bit, the mixing produces a 24-bit result from two original 12-bit numbers. At some point this has to be truncated, or rounded, back to 12-bits, either to fit within the structure of the equipment or to be accepted by external interfaces. Simply dropping the lower bits can result in unwanted visible artifacts, especially when handling pure, noise-free, computer generated pictures.

16 bits to 8 bits

See also: Digital mixing