**ISO/IEC 10918-1 : 1993(E)**

**F.2.1.3.1**

**Decoding model for DC coefficients**

The decoded difference, DIFF, is added to PRED, the DC value from the most recently decoded 8

×

8 block from the

same component. Thus ZZ(0)

=

PRED

+

DIFF.

At the beginning of the scan and at the beginning of each restart interval, the prediction for the DC coefficient is

initialized to zero.

**F.2.1.3.2**

**Decoding model for AC coefficients**

The AC coefficients are decoded in the order in which they occur in ZZ. When the EOB is decoded, all remaining

coefficients in ZZ are initialized to zero.

**F.2.1.4**

**Dequantization of the quantized DCT coefficients**

The dequantization of the quantized DCT coefficients as described in Annex A, is accomplished by multiplying each

quantized coefficient value by the quantization table value for that coefficient. The decoder shall be able to use up to four

quantization table destinations.

**F.2.1.5**

**Inverse DCT (IDCT)**

The mathematical definition of the IDCT is given in A.3.3.

After computation of the IDCT, the signed output samples are level-shifted, as described in Annex A, converting the

output to an unsigned representation. For 8-bit precision the level shift is performed by adding 128. For 12-bit precision

the level shift is performed by adding 2 048. If necessary, the output samples shall be clamped to stay within the range

appropriate for the precision (0 to 255 for 8-bit precision and 0 to 4 095 for 12-bit precision).

**F.2.2**

**Baseline Huffman Decoding procedures**

The baseline decoding procedure is for 8-bit sample precision. The decoder shall be capable of using up to two DC and

two AC Huffman tables within one scan.

**F.2.2.1**

**Huffman decoding of DC coefficients**

The decoding procedure for the DC difference, DIFF, is:

T

=

DECODE

DIFF

=

RECEIVE(T)

DIFF

=

EXTEND(DIFF,T)

where DECODE is a procedure which returns the 8-bit value associated with the next Huffman code in the compressed

image data (see F.2.2.3) and RECEIVE(T) is a procedure which places the next T bits of the serial bit string into the low

order bits of DIFF, MSB first. If T is zero, DIFF is set to zero. EXTEND is a procedure which converts the partially

decoded DIFF value of precision T to the full precision difference. EXTEND is shown in Figure F.12.

**104**

**CCITT Rec. T.81 (1992 E)**