HOW JPEG WORKS?

JPEG standard specifies 4 modes of operation: sequential DCT-based, progressive DCT-based, lossless and hierarchical. Here we will only focus on the DCT mode of operation.

Essentially, the generic DCT-based coding system employed in all the lossy image and video compression standards comprises the following steps:

• 8x8 Discrete Cosine Transform (DCT) - a spatial to frequency domain transformation.
• Quantization - Discard unimportant DCT Domain Samples.
• Entropy Coding - Lossless coding of quantized DCT outputs using Huffman or Arithmetic Coding.

First, the image is divided into blocks of 8x8 pixels. Each block is transformed into the frequency domain by a 2-D DCT.

The DCT transformation decomposes each input block into a series of waveforms, each with a particular spatial frequency. There are 64 waveforms composing the DCT basis functions. The DCT transformation can be viewed as computing for each waveform the corresponding weight so that the sum of the 64 waveforms scaled by the corresponding weights yield the reconstructed version of the original 8x8 matrix.

Then the DCT ouput coefficients are quantized and entropy coded.

USEFUL LINKS

• JPEG related links
• SPIHT Image Compression