ISO/IEC 10918-1 : 1993(E)
G.1.3.3 Coding model for subsequent scans of successive approximation
The procedure "Encode_AC_Coefficient_SA" shown in Figure G.10 increases the precision of the AC coefficient values
in the band by one bit.
As in the first scan of successive approximation for a component, an EOB decision is coded at the start of the band and
after each non-zero coefficient.
However, since the end-of-band index of the previous successive approximation scan for a given component, EOBx, is
known from the data coded in the prior scan of that component, this decision is bypassed whenever the current index, K,
is less than EOBx. As in the first scan(s), the EOB decision is also bypassed whenever the last coefficient in the band is
not zero. The decision ZZ(K)
=
0 decodes runs of zero coefficients. If the decoder is at this step of the procedure, at least
one non-zero coefficient remains in the band of the block being coded. If ZZ(K) is not zero, the procedure in Figure G.11
is followed to code the value.
The context-indices in Figures G.10 and G.11 are defined in Table G.2 (see G.1.3.3.1). The signs of coefficients with
magnitude of one are coded with a fixed probability value of approximately 0.5 (Qe = X'5A1D', MPS
=
0).
G.1.3.3.1
Statistical model for subsequent successive approximation scans
As shown in Table G.2, each statistics area for subsequent successive approximation scans of AC coefficients consists of a
contiguous set of 189 statistics bins. The signs of coefficients with magnitude of one are coded with a fixed probability
value of approximately 0.5 (Qe
=
X'5A1D', MPS
=
0).
G.2
Progressive decoding of the DCT
The description of the computation of the IDCT and the dequantization procedure contained in A.3.3 and A.3.4 apply to
the progressive operation.
Progressive decoding processes must be able to decompress compressed image data which requires up to four sets of
Huffman or arithmetic coder conditioning tables within a scan.
In order to avoid repetition, detailed flow diagrams of progressive decoder operation are not included. Decoder operation
is defined by reversing the function of each step described in the encoder flow charts, and performing the steps in reverse
order.
CCITT Rec. T.81 (1992 E)
129