The Still Image Lossy Compression Standard - JPEG

Overview Of The Jpeg Algorithm

JPEG compression algorithm consists of several steps. The steps are summarized in Figure 1. In this section the steps will be comprehensively explained.

Figure 1.

JPEG model for a lossy image compression

The first step transforms the image color into a suitable color space. There are several methods to transform the image into a color space (Hearn & Baker, 1986), (Jain, 1986). The most common methods are the split into YUV components (Hunt, 1995) or the split into RGB components (Laplante & Stoyenko, 1996). These components are interleaved together within the compressed data. The ratio between these components is usually not one to one. When YUV components are used, usually the Y component will have a four times weight. The human eye is less sensitive to the frequency of chrominance information than to the frequency of luminance information which is represented by the Y component in the YUV format. Hence, the Y component gets a higher weight (Awcock, 1996).

JPEG employs Chroma subsampling which is a technique of encoding images by using less resolution for chrominance information than for luminance information, taking advantage of the human eye's lower sensitiveness for color differences than for luminance differences. JPEG supports the obvious 4:1:1 chroma subsampling which denotes the color resolution is quartered, compared to the luminance information i.e. for each sampled element as in figure 2, there is 4 numbers for luminance and just one number for chrominance; however the default chroma subsampling of JPEG is 4:2:0 which denotes the horizontal sampling is doubled compared to 4:1:1, but as the U and V components are only sampled on each alternate line.