1 #+TITLE: ImageSqueeze - lossy image codec
2 * (document settings) :noexport:
3 ** use dark style for TWBS-HTML exporter
4 #+HTML_HEAD: <link href="https://bootswatch.com/3/darkly/bootstrap.min.css" rel="stylesheet">
5 #+HTML_HEAD: <script src="https://cdnjs.cloudflare.com/ajax/libs/jquery/1.11.3/jquery.min.js"></script>
6 #+HTML_HEAD: <script src="https://cdnjs.cloudflare.com/ajax/libs/twitter-bootstrap/3.3.5/js/bootstrap.min.js"></script>
7 #+HTML_HEAD: <style type="text/css">
8 #+HTML_HEAD: footer {background-color: #111 !important;}
9 #+HTML_HEAD: pre {background-color: #111; color: #ccc;}
13 - This program is free software: released under Creative Commons Zero
18 - Homepage: https://svjatoslav.eu
19 - Email: mailto://svjatoslav@svjatoslav.eu
21 - [[https://www.svjatoslav.eu/projects/][Other software projects hosted at svjatoslav.eu]]
24 - [[https://www2.svjatoslav.eu/gitweb/?p=imagesqueeze.git;a=snapshot;h=HEAD;sf=tgz][Download latest snapshot in TAR GZ format]]
26 - [[https://www2.svjatoslav.eu/gitweb/?p=imagesqueeze.git;a=summary][Browse Git repository online]]
28 - Clone Git repository using command:
29 : git clone https://www2.svjatoslav.eu/git/imagesqueeze.git
32 Lossy image codec. Optimized for photos. I developed it to test out an
33 image compression ideas.
35 I believe my [[id:1d917f74-e763-4a71-976e-4aa60732efa6][algorighm has interesting advantages.]]
37 Below are original photo and the same image being compressed down to
38 ~93 Kb and then decompressed.
40 [[file:originalAndCompressed.png]]
42 When looking very closely, slight grainyness, loss of color precision
43 and blurriness (loss of detail) could be noticed as a compression
44 artifacts. Still sharp edges are always preserved. Also no blocks
45 typical to JPEG are ever seen. I think that is awesome result for
46 just ~ 2.5 bits per pixel on that color photo.
47 * Algorithm description
48 + Color image is split into three separate channels: Y, U and V.
49 + Each channel is independently compressed.
51 + Single channel compression method:
53 + Algorithm is inspired by [[https://en.wikipedia.org/wiki/Diamond-square_algorithm][diamond-square algorithm]] that is meant to
54 produce random heightmap/plasma effect.
56 + During image compression: 2D image surface is iterated in the
57 similar manner to diamond-square algorithm. Average color from
58 neighbors is calculated and difference between neighbors average
59 and actual pixel color is saved. As the pixel grid becomes
60 gradually more dense, difference between neighbors tends to get
61 smaller, thereby requiring less bits per pixel for storing the
63 ** Algorighm advantages
65 :ID: 1d917f74-e763-4a71-976e-4aa60732efa6
67 + It can be applied to any amount of dimensions, even for sound and
70 + Algorithm can operate in lossy and lossless mode.
72 + Algorithm naturally handles progressive loading. That is: low
73 resolution thumbnail of entire thing is immediately available and
74 gets gradually more dense during entire loading process.
76 + Algorithm naturally supports variable resolution. That is: different
77 areas can be encoded with different resolutions / pixel densities.
79 + Fast: Very little computations per pixel.
81 * TODO Things to improve
82 - Code documentation is weak.
84 - Better sample applications needed:
85 - Commandline image conversion utility.
88 - Add lossless support.
90 - Extract algorithm key parts into reusable library and make it
91 variable dimensional. So that the same code can be used for sound,
92 image and volumetric data compression/decompression.