Digital Signal Processing Reference
In-Depth Information
14.7.3 JPEG Color Image Compression
This section is devoted to reviewing JPEG standard compression and examines the steps briefly. We
focus on the encoder, since the decoder is just the reverse process of the encoding. The block diagram
for the JPEG encoder is in Figure 14.39 .
The JPEG encoder has the following main steps:
1. Transform RGB to YIQ or YUV (U and V ΒΌ chrominance components).
2. Perform DCT on blocks.
3. Perform quantization.
4. Perform zigzag ordering, DPCM, and run-length encoding.
5. Perform entropy encoding (Huffman coding).
RGB to YIQ Transformation
The first transformation is of the RGB image to a YIQ or YUV image. Transformation from RGB to
YIQ has previously been discussed. The principle is that in YIQ format, the luminance channel carries
more signal energy, up to 93%, while the chrominance channels carry up to only 7% of signal energy.
After transformation, more effort can be spent on coding the luminance channel.
DCT on Image Blocks
Each image is divided into 8 8 blocks. 2D-DCT is applied to each block to obtain the 8 8 DCT
coefficient block. Note that there are three blocks, Y, I, and Q.
Quantization
The quantization is operated using the 8 8 quantization matrix. Each DCT coefficient is quantized,
that is, divided by the corresponding value given in the quantization matrix. In this way, a smaller
RGB
YIQ or YUV
8x8
8x8
8x8
Quantization
DCT
p ij
F uv
(, )
(,)
Quantization
tables
Coding
Tables
Zigzag
DC
Header
Tables
DPCM
Entropy
coding
RLC
Data
AC
FIGURE 14.39
Block diagram for JPEG encoder.
 
Search WWH ::




Custom Search