sparse image representation via combined transforms - Convex ...
sparse image representation via combined transforms - Convex ...
sparse image representation via combined transforms - Convex ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
42 CHAPTER 3. IMAGE TRANSFORMS AND IMAGE FEATURES<br />
Fralick developed a fast algorithm that would take N log 2 N − 3N/2 + 4 multiplications and<br />
3N/2(log 2 N − 1) + 2 additions; when N = 8, their algorithm would take 16 multiplications<br />
and 26 additions. This is a significant reduction in complexity. In 1984, Wang [139] gave an<br />
algorithm that would take N(3/4log 2 N − 1) + 3 multiplications and N(7/4log 2 N − 2) + 3<br />
additions; when N = 8, this is 13 multiplications and 29 additions. Also in 1984, Lee [92]<br />
introduced an algorithm that would take N/2log 2 N multiplications and 3N/2log 2 N −N +1<br />
additions; when N = 8, this is 12 multiplications (one less than 13) and 29 additions.<br />
Algorithm # of Multiplications # of Additions<br />
Definition N 2 N(N − 1)<br />
{64} {56}<br />
Chen, Smith and N log 2 N − 3N/2+4 3N/2(log 2 N − 1) + 2<br />
Fralick, 1977 [28] {16} {26}<br />
Wang, 1984 [139] N(3/4log 2 N − 1) + 3 N(7/4log 2 N − 2) + 3<br />
{13} {29}<br />
Lee, 1984, [92] N/2log 2 N 3N/2log 2 N − N +1<br />
{12} {29}<br />
Duhamel, 1987 [62] 2N − log 2 N − 2 Not Applicable<br />
(theoretical bound) {11}<br />
Table 3.3: Number of multiplications and additions for various fast one-D DCT/IDCT<br />
algorithms. The number in {·}is the value when N is equal to 8.<br />
The overall complexity of the DCT arises from two parts: multiplicative complexity<br />
and additive complexity. The multiplicative complexity is the minimum number of nonrational<br />
multiplications necessary to perform DCTs. The additive complexity, correspondingly,<br />
is the minimum number of additions necessary to perform DCTs. Since it is more<br />
complex to implement multiplications than additions, it is more important to consider multiplicative<br />
complexity. In 1987, Duhamel [62] gave a theoretical bound on the 1-D N-point<br />
DCT: it takes at least 2N − log 2 N − 2 multiplications. In 1992, Feig and Winograd [67]<br />
extended this result to an arbitrary dimensional DCT with input sizes that are powers<br />
of two. Their conclusion is that for L-dimensional DCTs whose sizes at coordinates are<br />
2 m 1<br />
, 2 m 2<br />
,... ,2 m L,withm 1 ≤ m 2 ≤ ...m L , the multiplicative complexity is lower bounded<br />
by 2 m 1+m 2 +...+m L−1<br />
(2 mL+1 − m L − 2). When L = 1, this result is the same as Duhamel’s<br />
result [62].<br />
In <strong>image</strong> coding, particularly in JPEG, 2-D DCTs are being used. In 1990, Duhamel