备忘

1. 构造的小波应该写成((1+z)/2)^p*R(z)的形式，可以对消失矩一目了然
2. 关于计算复杂度的讨论。由于D9/7的提升系数是无理数，必然是有限截断的有理数表示，在实际计算中采用先把系数左移13位，在整数乘法运算，得到的结果右移13位。即s(l)=s(l)+[(a1*2^13)*(d(l-1)+d(l))/2^13]，而且有些程序仅保留a1到小数点后4位，精度很差。相比s(l)=s(l)+(d(l-1)+d(l))>>3这样的运算，要复杂得多。

3. discussion on complexity should have data like the computing time consuming etc. to prove it, so it seems to have more precise.
4. D9/7 wavelets filters can be written as the form in Taubman's book, so it has the none rational form.

5. in fact D9/7 is an unreversiable wavelets, so interger version of its lifting scheme(the same with the fix-points version) is finite precise realization, specially the scale process is unreversiable.  D5/3 is a reversiable wavelets. The wavelets we compared must restrict in both types, reversiable and unreversible.

6. 值得注意的是＝it is important to note that
7.someone say " reversible integer-to-integer, and nonreversible real-to-real transforms".

In fact, D9/7 is a nonreversible real-to-real transforms, but some is a reversible integer-to-integer, but in lossy case, its performance approachs...

8.computational complexity = 计算量
9.there should be description on the structure of paper in the introduction.

10. more theorical, increase the depth
11. make the introduction systematic, refer more work previous
12. add line-base, and states figure
13. show that we have done a lot work...