aboutsummaryrefslogtreecommitdiff
path: root/packages/noncvs/windows/extra/Gem/manual/Pixes.html
blob: badf8bc9461f99267b7baf3542d2880bef21eefb (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
<!doctype html public "-//w3c//dtd html 4.0 transitional//en">
<html>
<head>
  <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-1">
  <meta name="Author" content="Mark Danks">
  <meta name="Author" content="IOhannes m zmölnig">
  <title>Pixes (image processing)</title>
</head>
<body>

<center>
<h2>
<u>Image processing</u></h2></center>
The pix objects are used to do image processing to pixel data. If
you load in an image with <i>[pix_image]</i>, then you can change what the
image looks like before rendering it out
<p>In general, processing images is <i>extremely</i> 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.<br>
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.

<p>The pix objects are divided into two general groups, those which take
one input, and those which require two input images. For example,
<i>[pix_invert]</i>
will "invert" all of the pixels (if a pixel is white, it will change to
black), while <i>[pix_add]</i> will add two images together.
<p>Only some of the pix objects are described here. Look in the reference
patches for explanations for the other pix objects.
<p><a href="#invert">[pix_invert]</a> - invert the pixel data
<br><a href="#add">[pix_add]</a> - add two pixes together
<br><a href="#mask">[pix_mask]</a> - create an alpha mask
<br><a href="#convolve">[pix_convolve]</a> - convolve a pix with a kernel
<p><img SRC="tribar.gif" height=13 width=561>
<h3>
<a NAME="invert"></a>[pix_invert]</h3>
<i>[pix_invert]</i> inverts the pixels in an image. To use <i>[pix_invert]</i>,
simply make sure that you have already loaded an image into the chain. 
In the following patch, the fractal image will be inverted.
<center>
<p><img SRC="invert.jpg" BORDER=1 height=120 width=179></center>

<p>Here is the difference between the fractal image and the inverted version.
<center>
<p><img SRC="normalFrac.jpg" height=256 width=256><img SRC="invertFrac.jpg" height=256 width=256></center>

<p><img SRC="tribar.gif" height=13 width=561>
<h3>
<a NAME="add"></a>pix_add</h3>
<i>[pix_add]</i> does what you would expect. It adds two images together.
<center>
<p><img SRC="add.jpg" BORDER=1 height=152 width=305></center>

<p>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.
<center>
<p><img SRC="addResult.jpg" height=257 width=255></center>

<p><br>
<p><img SRC="tribar.gif" height=13 width=561>
<h3>
<a NAME="mask"></a>pix_mask</h3>
<i>[pix_mask]</i> 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 <i>[alpha]</i>
object was removed, then you would just see the solid fractal image (because
the alpha channel wouldn't be used).
<p>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.

<i>[pix_mask]</i> only modifies the alpha channel and does not touch the
red, green, or blue data.
<center>
<p><img SRC="mask.jpg" BORDER=1 height=262 width=191></center>

<p>The result is this image.
<center>
<p><img SRC="maskResult.jpg" height=218 width=187></center>

<p><img SRC="tribar.gif" height=13 width=561>
<h3>
<a NAME="convolve"></a>pix_convolve</h3>
<i>[pix_convolve]</i> 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.
<p>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.
<p>Look at gem_pix/gemPixConvolve.pd to get an idea of some of the kernels
that you can send to <i>[pix_convolve]</i> and the effects that you can get.
<p>If you want to learn the math behind convolution, then find any standard
image processing (or audio processing book, this is just 2D convolution).
<br> 
<p><img SRC="tribar.gif" height=13 width=561>
<p><a href="index.html">[return]</a>
<br>
</body>
</html>