 |
Common features of lossless compression standard techniquesCompression of images is usually performed independently for each color component. Compression of RGB components is fully reversible. Typical compression ratios for monochromatic and color RGB natural images are around 1.2 - 2.5; depending on image content and used compression method. Conversion from RGB to other color space is possible but only reversible conversion is allowed. For example, the described earlier conversion to YCrCb is not reversible because of truncation errors.Using proper conversion to luminance / chrominance components and due to their properties - we can increase compression ratio by 20-50%. The described earlier intraframe prediction may be replaced by some other reversible operation, for example integer wavelet transform. Wavelet technique is explained in lesson 8. After technique selection, the coder produces bitstream of size dependent only on image content. There is no future possibility of compression ratio adaptation, so for two input images of the same size usually coder outputs streams of different sizes. Lossless mode in JPEGJPEG is the probably the best known image compression standard. Official name is: ISO/IEC IS 10918 "Information technology - digital compression and coding of continuous-tone still images". Algorithm used in JPEG standard for lossless coding is totally different from lossy coding mode described inthe next lesson. The lossless coding consists of two operations: nonadaptive intraframe prediction (using one of 7 predictors presented earlier this lesson) followed by entropy coding: Huffman or arithmetic. Huffman coding is the base mode. Arithmetic coding is an option mode, needs more operations, was patented and usually increases a compression ratio (comparing to Huffman coding) only of few percents (2.5% at average).
|