Image Processing Basics - Digital Signal Processing: Fundamentals and Applications - page 738

FIGURE 14.40
JPEG compressed color image.
Symbol 1: (run length, size)
Symbol 2: (amplitude)
The 4-bit run length can only tackle runs for zeros from 1 to 15. If the run length of zeros is larger than
15, then a special code (15, 0) is used for Symbol 1. Symbol 2 is the amplitude in Huffman coding as
shown in Table 14.14 , while the encoded Symbol 1 is kept in its format:
(run length, size, amplitude)
Let us code the following run-length code of AC coefficients:
ð 0 ; 1 Þ; ð 1 ; 1 Þ; ð 2 ; 1 Þ; ð 0 ; 3 Þ; ð 3 ; 2 Þ; ð 0 ; 0 Þ
We can produce a bit stream for AC coefficients:
ð 0000 ; 0001 ; 0 Þ; ð 0001 ; 0001 ; 0 Þ; ð 0010 ; 0001 ; 0 Þ;
ð 0000 ; 0010 ; 11 Þ; ð 0011 ; 0010 ; 10 Þ; ð 0000 ; 0000 Þ
There are 55 bits in total. Figure 14.40 shows a JPEG compressed color image (included in the color
insert). The decompressed image is indistinguishable from the original image after comparison.
14.7.4 Image Compression Using Wavelet Transform Coding
We can extend the one-dimensional discrete wavelet transform (DWT) discussed in Chapter 13 to the
two-dimensional DWT. The procedure is described as follows. Given an N N image, the 1D-DWT
using level 1 is applied to each row of the image; and after all the rows are transformed, the level-1 1D-
DWT is applied again to each column. After the rows and columns are transformed, we achieve four
first-level subbands labeled LL, HL, LH, and HH as shown in Figure 14.41 (a). The same procedure
repeats for the LL band only and results in the second-level subbands: LL2, HL2, LH2, and HH2