Byakuren

A theme color extracting library implemented by C.

This library is used in a related company for years.

TODO: Using KD-Tree to find some colors.

聖 白蓮

聖 白蓮（ひじり びゃくれん，Hiziri Byakuren）是系列作品《东方project》中的角色，首次登场于《东方星莲船》。

• 种族：魔法使
• 能力：使用魔法程度的能力（酣畅增强身体能力的）
• 危险度：不明
• 人类友好度：中
• 主要活动场所：命莲寺之类

命莲寺的住持。虽然原本是人类，不过由于常年的修行已经完全超越了人类。现在已经属于人们常说的魔法使了。

虽然已经入了佛门，但是不知道什么原因却被妖怪敬仰着。她从来没有像童话故事中的魔法使那样，念诵着咒语治退妖怪。使用的力量完全是邪恶的，一点都不像是圣人，虽然并没有人目击到她与人类为敌，但其实已彻底成为妖怪的同伴了。

Compile Static Library

Clone the project first.

$ git clone --recurse-submodules https://github.com/XadillaX/byakuren.git

This project can be compiled to a static libary (byakuren.a) for using.

$ make byakuren

After compiling, you may use this library just by including byakuren.h in your project.

#include "byakuren.h"

Supported Algorithm

• octree algorithm
• min-diff algorithm
• mix-in algorithm

APIs

bkr_rgb Structure

typedef struct bkr_rgb {
    uint8_t red;
    uint8_t green;
    uint8_t blue
} bkr_rgb;

RGB pixel structure.

name	type	description
red	uint8_t	the RED value (0-255)
green	uint8_t	the GREEN value (0-255)
blue	uint8_t	the BLUE value (0-255)

bkr_color_stats Structure

typedef struct bkr_color_stats {
    bkr_rgb color;
    uint32_t value;
    uint32_t count;
} bkr_color_stats;

Stats of theme color result.

name	type	description
color	bkr_rgb	a color pixel to indicate a theme color
value	uint32_t	a color pixel's INT32 value to indicates the theme color
count	uint32_t	stats of this theme color in the picture

typedef struct bkr_palette_array {
    uint32_t count;
    bkr_rgb* colors;
} bkr_palette_array;

A crowd of colors to indicate a theme color palette.

name	type	description
count	uint32_t	color count in this palette
colors	bkr_rgb*	each color in this palette

bkr_mindiff_parameter Structure

typedef struct bkr_mindiff_parameter {
    bkr_palette_array* palette;
    int16_t gray_offset;
} bkr_mindiff_parameter;

A parameter passes to Min-diff Algorithm.

name	type	description
palette	bkr_palette_array*	a palette to calculate the theme color, left for NULL to indicate the default palette
gray_offset	int16_t	the offset to judge whether a color is gray, recommand to be 5

Initialization & Release

Before the whole work you should initialize the Byakuren environment:

int bkr_init();

And after all the work you should release the environment:

void bkr_destroy();

Octree Algorithm

Build Octree

bkr_octree_node* bkr_build_octree(
        bkr_rgb* pixels,
        uint32_t pixel_count,
        uint32_t max_colors);

parameter	type	description
pixels	bkr_rgb*	the RGB pixels of a picture
pixel_count	uint32_t	pixel count of the picture
max_colors	uint32_t	maximum theme color count this octree will have

• Return an octree

Calculate

int bkr_octree_calculate_color_stats(
        bkr_octree_node* node,
        bkr_color_stats stats[]);

parameter	type	description
node	bkr_octree_node*	the octree which bkr_build_octree returned
stats	bkr_color_stats	an array to receive each theme color stats

• Return the count of theme colors.

Release Octree

void bkr_release_octree(bkr_octree_node* node);

parameter	type	description
node	bkr_octree_node*	the octree to be released

Example

bkr_rgb* rgb = GET_PICTURE_RGB(); // implement by yourself
uint32_t color_count = GET_PICTURE_PIXEL_COUNT(); // implement by yourself
bkr_color_stats stats[256];
bkr_octree_node* root = bkr_build_octree(rgb, color_count, 256);
int colors = bkr_octree_calculate_color_stats(root, stats);

Min-diff Algorithm

Calculate

int bkr_mindiff_calculate_color_stats(
        bkr_rgb* pixels,
        uint32_t pixel_count,
        bkr_color_stats stats[],
        bkr_mindiff_parameter* param);

parameter	type	description
pixels	bkr_rgb*	the RGB pixels of a picture
pixel_count	uint32_t	pixel count of the picture
stats	bkr_color_stats	an array to receive each theme color stats
param	bkr_mindiff_parameter*	the parameter passes to Min-diff Algorithm for calculating

• Return the count of theme colors.

Example

bkr_rgb* rgb = GET_PICTURE_RGB(); // implement by yourself
uint32_t color_count = GET_PICTURE_PIXEL_COUNT(); // implement by yourself
bkr_color_stats stats[256];
bkr_mindiff_parameter param;
param.gray_offset = 5;
param.palette = NULL;
int colors = bkr_mindiff_calculate_color_stats(rgb, color_count, stats, &param);

Mix Algorithm

Mix Octree and Min-diff up.

Calculate

int bkr_mix_calculate_color_stats(
        bkr_rgb* pixels,
        uint32_t pixel_count,
        uint32_t octree_max_colors,
        bkr_mindiff_parameter* mindiff_param,
        bkr_color_stats stats[]);

parameter	type	description
pixels	bkr_rgb*	the RGB pixels of a picture
pixel_count	uint32_t	pixel count of the picture
octree_max_colors	uint32_t	maximum theme color count this octree will have
param	bkr_mindiff_parameter*	the parameter passes to Min-diff Algorithm for calculating
stats	bkr_color_stats	an array to receive each theme color stats

• Return the count of theme colors.

Example

bkr_rgb* rgb = GET_PICTURE_RGB(); // implement by yourself
uint32_t color_count = GET_PICTURE_PIXEL_COUNT(); // implement by yourself
bkr_color_stats stats[256];
bkr_mindiff_parameter param;
param.gray_offset = -1;
param.palette = NULL;
int colors = bkr_mix_calculate_color_stats(rgb, color_count, 256, &param, stats);

Test Command

$ make ./test/bkr_test

After make command, you should generate a binary file named test.rgb. Then run:

$ cd test && ./bkr_test ALGORITHM

You may create your own .rgb file by referring test/run.js or test/test.c. Or you may have a look at Test Helper.

ALGORITHM is a parameter means algorithm you want to test.

support octree、mindiff and mix so far.

Test Helper

If you want to test quickly (no *.rgb), you may use a simple script.

Install Node.js first and come to test folder. Then run:

$ cd test
$ npm install

NOTICE: You can start test helper directly under OSX. Otherwise, you should compile a test binary executor before testing, that is $ make ./test/bkr_test.

After installing, run --help.

$ ./run.js --help

You will see some introduction. -u means a URL.

For an example:

$ ./run.js -u http://cdn.duitang.com/uploads/item/201205/22/20120522224448_43nFu.thumb.600_0.jpeg -a octree

After command above, a browser will be open to display the result.

If you want to test three algorithm at one time, you should make -a be all.

For an example:

$ ./run.js -u http://cdn.duitang.com/uploads/item/201205/22/20120522224448_43nFu.thumb.600_0.jpeg -a all

How It Look Like

With octree_max_colors be 16 in Octree and Mix algorithm.

Contribution

You're welcome to make Pull Requests.

「雖然我覺得不怎麼可能有人會關注我」