| Where is the error in this DCT-approach? [message #68411] |
Tue, 27 October 2009 14:02  |
rogass
Messages: 200
Registered: April 2008
|
Senior Member |
|
|
Hi there,
I cant find the error in this little program to compute forward and
backward the discrete cosine transform of an image. It seems to be
also difficult to avoid the loops. Please help me. Here is the code:
function cr_dct,dat,mode,shiftit=shiftit
sz = size(dat,/dimensions)
dim = n_elements(sz)
mode = n_elements(mode) eq 0? -1 : -2>mode<2
shiftit = keyword_set(shiftit)? 1 : 0
case dim of
1 : begin
end
2: begin
x = sz[0]
y = sz[1]
c = fltarr(sz)
sqrtx = sqrt(x)
sqrty = sqrt(y)
sqrt2 = sqrt(2)
sqrt2x = sqrt2/sqrtx
sqrt2y = sqrt2/sqrty
arr = c
indarr = (((l=findgen(x>y)))#(2.*l+1))*!dtor
cosxarr = cos(indarr/(2.*x))
cosyarr = x eq y? cosxarr : cos(indarr/(2.*y))
for u=0,x-1 do begin
for v=0,y-1 do begin
alphau = u eq 0 ? sqrtx : sqrt2x
alphav = v eq 0 ? sqrty : sqrt2y
tmparr = arr
;tmparr=dat*((r=rebin(cosxarr[u,*],x,y))) * $
; (x eq y? r: rebin(cosyarr[v,*],x,y))
;tmparr=(dat)[*] * ((r=(rebin(cosxarr[u,*],x,y))[*])) * $
; (x eq y? r : (rebin(cosxarr[v,*],x,y))[*])
for i=0,x-1 do $
for j=0,y-1 do $
tmparr[i,j] = mode eq -1? $
dat[i,j]*cosxarr[u,i]*cosyarr[v,j] : $
alphau*alphav*dat[i,j]*cosxarr[u,i]*cosyarr[v,j]
c[u,v] = mode eq -1? total(tmparr)*alphau*alphav : total(tmparr)
endfor
endfor
c = shiftit? shift(c,long(x/2),long(y/2)) : c
end
endcase
print,mode
return,c
end
pro test_dct
test=dist(16)
tvscl,congrid(test,128,128),0
tvscl, congrid(cr_dct(cr_dct(test,-1,/shiftit),1,/shiftit),128,128) ,1
end
Thank you and best regards
CR
|
|
|
|
| Re: Where is the error in this DCT-approach? [message #68558 is a reply to message #68411] |
Tue, 03 November 2009 02:36  |
rogass
Messages: 200
Registered: April 2008
|
Senior Member |
|
|
On 3 Nov., 00:44, pp <pp.pente...@gmail.com> wrote:
> On Nov 2, 9:27 pm, chris <rog...@googlemail.com> wrote:
>
>> Does nobody have any idea?
>
>> :(
>
>> CR
>
> You should at least describe the problem before expecting an answer.
> Not many people will try to interpret a page of code just to try to
> guess what problem you may have.
Ok, that's right. Thank you. The problem is that a forward followed by
a backward transformation of, for instance, a "dist"-image shows only
a simple cross and not the "dist"-image again. I cannot understand,
why the code won't work as expected. Please help me!
Thanks and best regards
CR
|
|
|
|
| Re: Where is the error in this DCT-approach? [message #68561 is a reply to message #68411] |
Mon, 02 November 2009 15:44 |
penteado
Messages: 866
Registered: February 2018
|
Senior Member
Administrator |
|
|
On Nov 2, 9:27 pm, chris <rog...@googlemail.com> wrote:
> Does nobody have any idea?
>
> :(
>
> CR
You should at least describe the problem before expecting an answer.
Not many people will try to interpret a page of code just to try to
guess what problem you may have.
|
|
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
