Pixes (image processing)

From 5c694f159b012d1d94bf15bfe580dec52e4fe797 Mon Sep 17 00:00:00 2001 From: Travis CI Date: Sat, 14 Mar 2015 20:40:49 +0000 Subject: Gem osx/x86_64 built '' for osx/x86_64 --- Gem/manual/Pixes.html | 105 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 105 insertions(+) create mode 100644 Gem/manual/Pixes.html (limited to 'Gem/manual/Pixes.html') diff --git a/Gem/manual/Pixes.html b/Gem/manual/Pixes.html new file mode 100644 index 0000000..badf8bc --- /dev/null +++ b/Gem/manual/Pixes.html @@ -0,0 +1,105 @@ + + + + + + + Pixes (image processing) + + + + +

+Image processing

+The pix objects are used to do image processing to pixel data. If +you load in an image with [pix_image], then you can change what the +image looks like before rendering it out +

In general, processing images is extremely expensive, so you +probably cannot have that many active pix objects. GEM only reprocesses +images when the source image changes or one of the parameters for a pix +object changes. This means that GEM will only process an image when +something is different, instead of every frame. If you want to do +a lot of processing at start up, but then not change anything once the +patch is running, GEM will only do the computation once.
+Modern CPUs use SIMD (Single Instruction - Multiple Data) (like MMX, SSE2, altivec) +to make pixel-processing more effective (by processing data parallely). +Until now, only the macOS version of Gem has support for SIMD for some pix-objects. +MMX/SSE2 boosts will hopefully come in future Gem-releases. + +

The pix objects are divided into two general groups, those which take +one input, and those which require two input images. For example, +[pix_invert] +will "invert" all of the pixels (if a pixel is white, it will change to +black), while [pix_add] will add two images together. +

Only some of the pix objects are described here. Look in the reference +patches for explanations for the other pix objects. +

[pix_invert] - invert the pixel data +
[pix_add] - add two pixes together +
[pix_mask] - create an alpha mask +
[pix_convolve] - convolve a pix with a kernel +

+[pix_invert]

+[pix_invert] inverts the pixels in an image. To use [pix_invert], +simply make sure that you have already loaded an image into the chain. +In the following patch, the fractal image will be inverted. + +

+ +

Here is the difference between the fractal image and the inverted version. +

+ +

+pix_add

+[pix_add] does what you would expect. It adds two images together. + +

+ +

This patch adds the fractal image with a car image. The processed +image will often contain a lot of white pixels, because the data is just +added together. This occurs in the resulting image, shown below. +

+ +

+pix_mask

+[pix_mask] is used to create an alpha mask from another image. +In the following example (gem_pix/gemMaskDancer.pd), the fractal image's +alpha channel is replaced by the dancer image. If the [alpha] +object was removed, then you would just see the solid fractal image (because +the alpha channel wouldn't be used). +

In other words, images are composed of a red, a green, a blue, and an +alpha channel. The alpha channel is the transparency of the pixel. + +[pix_mask] only modifies the alpha channel and does not touch the +red, green, or blue data. +

+ +

The result is this image. +

+ +

+pix_convolve

+[pix_convolve] convolves pix data with a convolution kernel. +Basically, you can get really nice effects if you choose the correct kernel...and +garbage if you choose the wrong one. +

Edge detection is done with a convolution kernel, as is smoothing. +The biggest problem is that convolving an image is about the most expensive +operation that you can do in GEM. +

Look at gem_pix/gemPixConvolve.pd to get an idea of some of the kernels +that you can send to [pix_convolve] and the effects that you can get. +

If you want to learn the math behind convolution, then find any standard +image processing (or audio processing book, this is just 2D convolution). +
+

[return] +
+ + -- cgit v1.2.1