The number of calculations your phone or computer performs when displaying a JPEG image is astounding.
My rudimentary understanding, before watching this, was that areas of nearly identical pixels were reduced to a shape and a repeating color. The way it actually happens is incredibly complex and requires a lot more calculation than I realized, but relies on pattern matching and treating color with less resolution than brightness. Further explanation in this video illustrates why out eyes aren't good at detecting very high-frequency, low contrast portions of images, which can therefore be eliminated during the compression process to save space.
https://youtu.be/Ba89cI9eIg8
3:14 The method uses a standardized set of 64 patterns of light and dark to approximate an image. In the low-resolution image below, at the edges of the teenagers head are squares the size of his glasses where the algorithm chose a diagonal-chessboard pattern as an approximation of the few hairs sticking up.
No comments:
Post a Comment