&nbrs;Beyond Resolution

 


 [ethnography of DCT]

THE GUY BEHIND THE GUY BEHIND THE GUY*
<href: Cory Arcangel: on Compression, 2007>

A Discrete Cosine Transform or 64 basis functions of the JPEG compression (Joint Photographic Experts Group) 8 x 8 pixel macroblocks.



The .JPEG compression consists of these six subsequent steps
1. Color space transformation. Initially, the image has to be transformed from the RGB colorspace to
Y′CbCr. This colorspace consists of three components that are handled separately; the Y’ (luma or brightness) and the Cb and Cr values; the blue-difference and red-difference Chroma components.
2. Downsampling. Because the human eye doesn’t perceive small differences within the Cb and Cr space very well, these elements are downsampled, a process that reduces its data dramatically.
3. Block splitting. After the colorspace transformation and downsampling steps, the image is split into 8 x 8 pixel tiles or macroblocks, which are transformed and encoded separately.
4. Discrete Cosine Transform. Every Y’CbCr macroblock is compared to all 64 basis functions (base cosines) of a Discreet Cosine Transform. A value of resemblance per macroblock per base function is saved in a matrix, which goes through a process of reordering.
5. Quantization. The JPEG compression employs quantization, a process that discards coefficients with values that are deemed irrelevant (or too detailed) visual information. The process of quantization is optimized for the human eye, tried and tested on the Caucasian Lena color test card.
Effectively, during the quantization step, the JPEG compression discards most of all information within areas of high frequency changes in color (chrominance) and light (luminance), also known as high contrast areas, while it flattens areas with low frequency (low contrasts) to average values, by re-encoding and deleting these parts of the image data. This is how the rendered image stays visually similar to the original – least to human perception. But while the resulting image may look similar to the original, the JPEG image compression is Lossy, which means that the original image can never be reconstructed.
6. Entropy coding. Finally, a special form of lossless compression arranges the macroblocks in a zigzag order. A Run-Length Encoding (RLE) algorithm groups similar frequencies together while Huffman coding organizes what is left.
Revealing the surface and structure of the image *<href: Ted Davis: ffd8, 2012>
A side effect of the JPEG compression is that the limits of the images’ resolution – which involve not just the images’ number of pixels in length and width, but also the luma and chroma values, stored in the form of 8 x 8 pixel macroblocks – are visible as artifacts when zooming in beyond the resolution of the JPEG.
Because the RGB color values of JPEG images are transcoded into Y’CbCr macroblocks, accidental or random data replacements can result into dramatic discoloration or image displacement. Several types of artifacts can appear; for instance ringing, ghosting, blocking, and staircase artifacts. The relative size of these artifacts demonstrates the limitations of the JPEGs informed data: a highly compressed JPEG will show relatively larger, block-sized artifacts.