12月 04

python和C统计jpeg图片RGB值的效率

  测试硬件pcduino(arm),操作系统ubuntu 3.4.29+ armv7l GNU/Linux。测试文件,图片jpeg,大约1.3M。测试内容统计图片的RGB里的值,存在数据库里将来看看能不能做搜索用。
  先是用python,开发速度非常快,但是效率有点问题,所以有用C做了一下。以下是代码和两次的测试结果。C使用的jpeglib库8.0版本。

python代码

#!/bin/python
#-*- coding:utf-8 -*-
# Filename:   
# Revision:    1.0
# Date:        2013-12-04
# Author:      simonzhang
# web:         www.simonzhang.net
# Email:       simon-zzm@163.com
### END INIT INFO
from PIL import Image
# 对RGB简历字典
R = {}
G = {}
B = {}
# 初始化值
for i in xrange(256):
    R[i] = 0
    G[i] = 0
    B[i] = 0
# 大图片需要处理很长时间,所以要用python可以按比例将图缩小
a = Image.open('123.jpg')
#a = a.resize((int(a.size[0] * 0.1), int(a.size[1] * 0.1)), Image.ANTIALIAS)
x = a.size[0]
y = a.size[1]

# 循环处理每个像素点的RGB值
for lx in xrange(x):
    for ly in xrange(y):
        rgb = a.getpixel((lx, ly))
        R[rgb[0]] += 1
        G[rgb[1]] += 1
        B[rgb[2]] += 1
# 打印最终结果
print sorted(R.items(), key=lambda R:R[1],reverse=True)[0]
print sorted(G.items(), key=lambda G:G[1],reverse=True)[0]
print sorted(B.items(), key=lambda B:B[1],reverse=True)[0] 

C代码

/*
*Filename:   
* Revision:    1.0
* Date:        2013-12-04
* Author:      simonzhang
* web:         www.simonzhang.net
* Email:       simon-zzm@163.com
*/
#include 
#include 
#include 

/* we will be using this uninitialized pointer later to store raw, uncompressd image */
//unsigned char *raw_image = NULL;
/* */
int max_num;

/* find max values*/
int sort_values(int arr[255])
{
    int largest1, largest2, temp;
    max_num = 0;
    largest1 = arr[0];
    largest2 = arr[1];
    if (largest1 < largest2)
    {
        temp = largest1;
        largest1 = largest2;
        largest2 = temp;
    }
    int i;
    for (i = 2; i < 255; i++)
    {
        if (arr[i] >= largest1)
        {
            largest2 = largest1;
            largest1 = arr[i];
            max_num = i;
        }
        else if (arr[i] > largest2)
        {
            largest2 = arr[i];
        }
    }
    return largest1;
}


read_jpeg_file(char *filename)
{
    /* these are standard libjpeg structures for reading(decompression) */
    struct jpeg_decompress_struct cinfo;
    struct jpeg_error_mgr jerr;
    /* libjpeg data structure for storing one row, that is, scanline of an image */
    JSAMPROW row_pointer[1];
    FILE *infile = fopen( filename, "rb" );
    unsigned long location = 0;
    int i = 0;
    if ( !infile )
    {
        printf("Error opening jpeg file %s\n!", filename );
        return -1;
    }
    /* here we set up the standard libjpeg error handler */
    cinfo.err = jpeg_std_error( &jerr );
    /* setup decompression process and source, then read JPEG header */
    jpeg_create_decompress( &cinfo );
    /* this makes the library read from infile */
    jpeg_stdio_src( &cinfo, infile );
    /* reading the image header which contains image information */
    jpeg_read_header( &cinfo, TRUE );
    /* Uncomment the following to output image information, if needed. */
    /* Start decompression jpeg here */
    jpeg_start_decompress( &cinfo );
    /* allocate memory to hold the uncompressed image */
    //raw_image = (unsigned char*)malloc( cinfo.output_width*cinfo.output_height*cinfo.num_components );
    /* now actually read the jpeg into the raw buffer */
    row_pointer[0] = (unsigned char *)malloc( cinfo.output_width*cinfo.num_components );
    /* read one scan line at a time */
    int r[255], g[255], b[255];
    for(i=0;i<255;i++)
    {
        r[i] = 0;
        g[i] = 0;
        b[i] = 0;
    }
    while( cinfo.output_scanline < cinfo.image_height )
    {
        jpeg_read_scanlines( &cinfo, row_pointer, 1 );
        for( i=0; i

先是测试python部分
#time python countrgb.py

(12, 510858)
(17, 429677)
(9, 662996)

real 11m4.009s
user 10m42.200s
sys 0m1.090s

开始测试C的部分
首先优化编译一下
#gcc countrgb.c -l jpeg -O3 -o countrgb
#time ./countrgb 123.jpg
12,510858
17,429677
9,662996

real 0m0.750s
user 0m0.730s
sys 0m0.010s

  两次统计结果相同,说明统计方法没有问题。python用了11分钟,C用了0.75秒,看来真的不是一般的快了。C代码编译后的文件8K,静态编译500多K。不过我是喜欢用python,开发速度快,结构清晰。