anamorphing

    joumik
    Feb 28 2014 | 1:37 pm
    hi all ,
    i'll try to be clearer than last posts ...
    i'm trying to process some anamorphose of a picture .
    here is a patch , a heavy-CPU-loading one .
    i' m  unable to manage jit.expr (working on it but it's still too early) , gen codes or other GPU clever stuff , enough , to be fast and light .
    so i'm going the bad way , but this is enough for what i want to do .
    it doesn't work good ...
    a picture , on a videoplane , is fullscreened in a jit.window .
    a jit.gl.cornerpin is drawn on it (randomn points) , and the corner coordinates are being processed in order to produce a matrix that will be the "repos" matrix of a jit.repos .
    jit.repos' output is going to the jit.gl.cornerpin , as its texture .
    i should get a flat image , on all the jit.window ,
    but it's not the case ,and i can't figure out why .
    is the process good ( i mean "good") , but the use of the videoplane , not ?
    the process consists of a series of scale boxes , and  2 counters generating each pixel of the "repos matrix" , using these scaled values .
    any ideas why it's not working ?
    .... render is on "ortho 2" , cornerpin is on "cornermode relative" ,i don't know what else to say ...
    sorry for this quite messy patch , i have tried to make it the clearest i could .
     EDIT sorry , previous patch had a bug , this one is ok (2 times)
    cheers Matthieu
    • joumik
      Mar 01 2014 | 8:00 pm
      sorry to insist . i have read , in the forum , about jit.gl.cornerpin' s distortion . would anybody tell me if this is the reason why my problem is happening ?
      if i'm still not clear enough , please tell me , i'll try to explain differently .
      i have just realised that a note , in the patch , is not included , it was saying , at stage 3 "mapping pixels process" : "PLEASE , WAIT FOR ~ 10 sec. , AFTER CLICKING THIS BANG" .
      i' m sorry it went away ....
      Matthieu
    • joumik
      Mar 03 2014 | 4:05 pm
      i have it ! the distortions i'm getting are only happening between the videoplane (background) and the jit.gl.cornerpin instances. as soon as i try to put several jit.gl.cornerpin together , and remove the videoplane from the view , the process is working good , the seams totally match !
      it's been working for approximatively 4 days now , and i couldn' t see it , only looking at the distortion , compared with a videoplane ...
      so , that's a good start . i would now need to write a jit.expr formula ,in order to replace my so slow (but working) subpatch , that goes to the jit.repos right inlet . maybe , the best would even be a jit.gen code ...
      ...
      ...
      cheers ! Matthieu