Contents

1 PNM (Portable aNy Map) [7]

• A collection of lossless raw image formats invented by Jef Poskanzer in the 1980s.
• ASCII (suitable for e-mail, for example) and binary formats.
• Example (binary color images):

2 JPEG (Joint Photographic Experts Group) [4]

• Lossy image compression standard (ISO/IEC 10918) of 1992.
• Virtually lossless down to 1.0 bpp.
• Coding algorithm:
  1. RGB → YCbCr color space transform.
  2. Sub-sample Cb and Cr components.
  3. Split each component into 8 × 8 blocks.
  4. Transform each block using the DCT.
  5. Quantify the coefficients.
  6. Entropy compress the quantized coefficients. by means of RLE and Huffman Coding.
• Lossless [5, 1] and scalable extensions are also avaiable.

3 PNG (Portable Network Graphics) [6]

• Successor of the patented GIF (Graphics Interchange Format).
• PNG specification was authored by an ad-hoc group of computer graphics experts and enthusiasts, in 1996. In 2003 became a international standard (ISO/IEC 15948).
• Coding algorithm:
  1. Cross the pixels of the image in row-scan order, starting in the high-left corner.
  2. Use a linear predictor to compute a residual image.
  3. Encode the residual image using LZ77 (Lempel-Ziv 1977).

4 JPEG 2000 [3]

• ISO/IEC progressive (lossy-to-lossless) image compression standard (ISO/IEC 15444) created in 2000.
• Successor of JPEG.
• Coding algorithm:
  1. RGB → YCbCr color space transform.
  2. Sub-sample Cb and Cr components (if lossy compression).
  3. Compute the 2D-DWT (2-Dimensional Discrete Wavelet Transform).
  4. Encode the DWT coefficients, by bit-planes, using Binary Arithmetic Coding.

5 WebP [2]

• Developed by Google in 2010 (released under a BSD license).
• Derived from intra-coding mode of WebM.
• Coding algorithm:
  1. Divide the image into macroblocks of 16 × 16 pixels.
  2. RGB → YCbCr color space transform.
  3. Sub-sample Cb and Cr components (4:2:0).
  4. Predict the pixels of the block using the above, left and above-left pixels.
  5. Perform the DCT or the WHT of the residue codeblock.
  6. Encode these data using LZ77 + Huffman Coding.

References