BitJazz Approximating codecs produce a number of objectionable artifacts, some subtle, some obvious. Compression artifacts are flaws that show up in an image that shouldn't be there, and are there only because of flaws in the compression and decompression algorithms. Each time a movie is decompressed and recompressed with an approximating codec such as DV, JPEG, or MPEG, these artifacts get worse, an effect known as generational loss. Examples of common artifacts include SheerVideo, being a lossless video codec, never produces any artifacts.In the tables below, we show magnified examples of actual artifacts obtained by using professional-quality commercial JPEG and DV codecs. One of the most objectionable artifacts that approximating image codecs introduce is ghosting, also known as echoing, ringing. A particularly colorful name for ghosting is mosquito noise, so called because it's commonly seen along poorly compressed silhouettes of skin. In ghosting, high-contrast edges are flanked by parallel curves of alternating brightness. Here's a 4X magnified example where the diagonal edge from the original is flanked in the degraded image by light and dark ghosts that are echoes of the original edge.
In pixellation, also known as blocking, the image looks like it's divided up into big blocks or tiles, as in a mosaic or a cubist painting, with sharp discontinuities between block boundaries. In the subtle, actual-size example shown here, the smooth gradient of the original appears in the degraded image to be divided into four blocks of distinct shades of grey.
The least surprising artifact of approximating image compressors is blurring, fuzziness, or unsharpness. In blurring, fine image details become indistinct, as though seen through an unfocussed lens. In the unmagnified example shown here, the sharp text of the word "blurring" in the original becomes, in the degraded image, much harder to read or even to focus on. Note also the ghosting above and below the text, and the fading of the background.
Television audiences are so used to bleeding, also known as smearing, that they are generally completely unaware of this artifact. In bleeding, colors of contrasting hue or saturation bleed across sharp brightness boundaries, so that the colors appear to be sloppily painted on. The problem is worst acute in images with high color detail, such as blooming meadows. In the magnified example shown here, the red and yellow hues are smeared across the circular boundary, yielding oranges and browns. Note also the ghost circles in the center and the mosquito noise on the outside.
One of the worst artifacts resulting from the use of approximating image codecs is washout. In washout, a contrast in hue or saturation between colors having the same brightness (luma) is completely lost, both colors being replaced with an in-between color that may appear strikingly different from the original colors. In the 4X magnified example shown here, red and green hues of equal luma are washed out to a uniform brown.
|