Folks --- steps forward made, now for one more hurdle.
The note that started all this is at the bottom. After some effort ( "the sledgehammer approach" ) i have been able to create a core image kernel that has the exponent term hiding as data in the red color channel. Once a means is found to pull in this image, sort thru its data and send it ( or a form of it ) to the radius input of a gaussian blur patch, I will be done on this module.
The image I want blurred is elsewhere and can be fed to the image input on the gaussian blur patch.
So I have an image being created with data hanging out in the red channel. Joy filled this developers heart at
22:35 Eastern time and I decided to go home.
What I had planned on doing was feeding said image to a _javascript_ module, and pulling each pixels red data and passing that on to the radius input. This prompted 2 questions - One is what does an image look like in a _javascript_ module, so I can figure out how to access it and two - is it an array or is it a pixel?
I suppose from the last question I should really ask is there a better way to do this? I had conceived of mangling a core kernel to get more than the image output, so I could get a float output along side an image. This has proved unsuccessful to my limited knowledge of QC. Any clues from the choir?
Any help would be greatly appreciated. My thanks to all you folks!
----- Original Message ----
From: Daniel D'Errico
<email@hidden>
To: email@hidden
Sent: Tuesday, October 23, 2007 9:01:23 PM
Subject: Trying to twist up an adaptive gaussian blur
Agonies abound folks
The original MATLAB code has a gaussian blur whose value changes based on pixel, so I thought I could this by using a core image kernel to figure out the calculations, then feed the exponent of the E term to the radius input of the gaussian blur.
No such luck.
I can do the required calculations. I use image only to determine what value the euclidean distance map function will return. the division, rounding and multiplying all are more or less ok.
I cannot output the exp term. Kernels only output images.
so I turned my sorry fate over to a _javascript_ block but wait!! How will I get an image into
_javascript_ so I can get the coordinates of the day ( as samplercoords does in kernels )?
Thats my question. Is there any way either inside a _javascript_ block to determine the coords of the current pixel or outside a java block by having a patch send me the current coords?
any help is appreciated.
--dan
