源于大作业~~
目录
前言
一、实现算法
二、结果展示
三、算法框架
(1) QuadTreeNode.h
(2) 结点扩展、细化模糊层次
(3) 模糊化图像四叉树转为图像
(4) 主函数代码
四、说明
五、结语
六、震惊一百年
七、开源代码——but拒绝抄袭从你我做起
----------------------------------QuadTreeNode.h------------------------------------
----------------------------------QuadTreeFunc.h------------------------------------
----------------------------------QuadImg_todo.cpp---------------------------------
前言
一张图片常常会存在空间冗余,即一大部分区域的色彩值相同,然而存储时却将这些像素块视作不同的色彩值,以满足较好的格式规整化。除了存储图片外,对于图片精度不是很高的要求、或者说要求对图片进行一定的模糊化处理,这种种情况都要求对图片的空间冗余进行降低。
针对模糊处理图片的要求,对图片分成四个象限,用一颗四叉树记录,四叉树的叶子结点记录图片的像素值,中间结点用于评判/细分模糊层次。
这样就可以根据一张图片的冗余/细节的分布,自动地进行选择模糊层次。
关键词:四叉树,自适应模糊,图片处理,空间冗余
一、实现算法
模糊化/降低冗余等等,可以将一块区域的色彩值用一个RGB块代替,而不至于每个色彩都单独备份一个。而这个过程,可以将一块区域的像素值,通过计算、选择出一个代表的像素值,来代替这个区域的所有像素值。针对模糊图片问题,为了保持一定的平滑性,采取中位数/均值等简单方法可以做到,也可以进阶采用高斯模糊等等手段。我采用了简单的均值模糊。
然而,一张图片,总是有部分存在空间冗余,部分边缘细节较多,统一的进行模糊,是不理智且不满足要求的。对此,评判一张图片是否需要模糊化,我们可以通过判断其“离散”情况来选择。最简单的“离散”情况度量,即计算RGB三色的方差。当方差和阈值Tolerance进行比较,小于阈值意味着这一个区域的相似程度足够高可以进行均值模糊;大于阈值则意味着相似程度不够高,需要继续细分细化,再进行上述操作。
对此,可以得出以下递归算法:
①图像计算RGB三色方差D1,D2,D3。
②D1>Tolerance || D2>Tolerance || D3>Tolerance 跳转到③否则跳转到④
③标记子图为中间结点。将子图分为四个区域,对每个子图,跳转到①
④计算子图RGB的均值ER,EG,EB,将这一区域所有的像素RGB值赋为对应均值ER,EG,EB。标记为叶子结点。返回。
一开始,用一个四叉树结点root记录整张图片的像素信息,采用上述递归算法,可以将这一个root结点扩展到一棵树,这棵树的所有的叶子结点记录着整个模糊化后的图的信息。
将整棵树转化为图时,只需要遍历整棵树,将叶子结点的色彩区块信息取出并整合,就可以得到模糊化后的图片。
不是很明白算法?看下面这张图片你就明白了!
对图片划分越细,细节越多;反之越粗糙。对于那些分块后满足阈值要求的图像,就模糊并存储其像素值,否则继续细分——直到满足阈值或图片足够小。
二、结果展示
原图:
Tolerance:0
Tolerance:5
Tolerance:15
Tolerance:25
Tolerance:35
Tolerance:50
Tolerance:100(haha~~)
三、算法框架
因为大作业提交时间还没有截止,所以不方便开源所有代码,这里给出框架。
(1) QuadTreeNode.h
struct Pos //偏移量记录
{
int x, y;
};
struct color //像素RGB值记录
{
unsigned char r;
unsigned char g;
unsigned char b;
};
class QuadTreeNode
{
public:
//记录细分的四个子图(valid==false,需要细化的情况下链接子图)
QuadTreeNode *q1, *q2, *q3, *q4; // 1~4象限
//记录子图相对于原图的偏移量,以便树转图
Pos position;
//记录子图的长宽像素数
int height, width;
//记录是否为叶子结点,即是否为模糊化图像信息记录结点
bool valid;
//这一结点对应的范围的色彩信息
color **rgbs;
//记录结点深度
int depth;
public:
QuadTreeNode(color **r, int wd, int ht);
QuadTreeNode(int posx, int posy, int wd, int ht);
//计算是否差异超过Tolerance
bool VarianceCalculate(int Tolerance);
//随机化模糊
bool RandomPuzzyTag();
//计算R均值
int AverageR();
//计算G均值
int AverageG();
//计算B均值
int AverageB();
//模糊化区域
void Fuzzify();
};
(2) 结点扩展、细化模糊层次
//以Tolerance为模糊阈值,扩展rt结点(自然少不了递归)
void TreeFuzzifyExtend(QuadTreeNode *rt, int Tolerance){...}
(3) 模糊化图像四叉树转为图像
//将四叉树记录的模糊化后的图像像素信息传递给二维数组img[][]
void TreeToImage(QuadTreeNode *rt, color **img){...}
(4) 主函数代码
值得说明的是,实验所给的图片为ppm格式,因此读写比较特殊。
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include "QuadTreeNode.h"
// #include "QuadTree.h"
#include "QuadTreeFunc.h"
void printImage(char *fileName, int width, color **a);
// color color;
// To get ppm image from jpeg file, please visit https://convertio.co/jpg-ppm/
void readImage(int p, char *inFile, char *outFile) // Note that width == height here
{
FILE *f = fopen(inFile, "rb");
char u[3]; // placehoder
int width, height, max_value;
fscanf(f, "%s%d%d%d%c", u, &width, &height, &max_value, &u[0]);
int i;
color **colors, **img;
colors = (color **)malloc(height * sizeof(color *));
img = (color **)malloc(height * sizeof(color *));
for (i = 0; i < height; i++)
{
colors[i] = (color *)malloc(width * sizeof(color));
img[i] = (color *)malloc(width * sizeof(color));
}
for (i = 0; i < height; i++)
fread(colors[i], sizeof(color), width, f);
fclose(f);
//=============================================================
QuadTreeNode *rt = new QuadTreeNode(colors, width, height);
TreeFuzzifyExtend(rt,100);
TreeToImage(rt, img);
//=============================================================
printImage(outFile, width, img);
}
void printImage(char *fileName, int width, color **a) // Note that width == height here
{
FILE *f = fopen(fileName, "wb");
fprintf(f, "P6\n");
fprintf(f, "%d %d\n", width, width);
fprintf(f, "255\n");
int i;
for (i = 0; i < width; i++)
fwrite(a[i], sizeof(color), width, f);
fclose(f);
}
// int main(int argc, char **argv)
int main()
{
int tolerance = 0;
char inFile[100];
char outFile[100];
// if (argc > 1)
{
// tolerance = atoi(argv[1]);
// inFile = argv[2];
// outFile = argv[3];
strcpy(inFile,"D:\\...\\a.ppm");
strcpy(outFile,"D:\\...\\result.ppm");
readImage(tolerance, inFile, outFile);
}
return 0;
}
除了正常的I/O和预处理,自己编写的部分,千言万语汇成主函数的几行话:
四、说明
1、可以采用命令行读取指令的方式,但是为了方便调试,我修改了那部分源代码。可以通过重载的main函数 int main(int argc, char **argv) 来实现。
2、除了均值模糊外,还可以选择中位数模糊
3、随机模糊化处理的部分,被我注释掉了
4、将ppm格式转为jpg格式可以简单的采用python库函数
from PIL import Image
img = Image.open("D:\\...\\a.ppm")
img.save("D:\\...\\a.jpg")
#img.show()
5、为了得到更加平滑的处理结果,可以了解一下高斯模糊这种高级的玩意
6、其余源代码将在大作业提交结束后上传开源。
五、结语
无论数据结构与算法还是程序设计等等 ,不要局限于作业,做一些小东西,小项目,小工程,小工具等等,都会让学习变得有趣。谁不想有一个能吹牛逼的本科呢?
=========================================================================
六、震惊一百年
今天打开博客,
什么玩意?
本蒟蒻萌新首被挂名,感慨激动万分——同学们的做作业热情太高涨了!
感觉不开源都对不起这个情况(流汗黄豆)——骑虎难下,亚历山大文章来源:https://www.toymoban.com/news/detail-461384.html
七、开源代码——but拒绝抄袭从你我做起
感觉不开源都对不起这个热度了。——但是提供思路与个人细节,具体修缮、编写代码还靠个人,抄袭打灭,对你对我都不好~~文章来源地址https://www.toymoban.com/news/detail-461384.html
----------------------------------QuadTreeNode.h------------------------------------
#ifndef __QUADTREENODE_H__
#define __QUADTREENODE_H__
#ifndef NULL
#define NULL 0
#endif
#include <cmath>
#include <iostream>
#include <random>
using namespace std;
struct Pos
{
int x, y;
};
struct color
{
unsigned char r;
unsigned char g;
unsigned char b;
};
class QuadTreeNode
{
public:
QuadTreeNode *q1, *q2, *q3, *q4; // 1~4象限
Pos position;
int height, width;
bool valid;
color **rgbs;
int depth;
public:
QuadTreeNode(color **r, int wd, int ht);
// QuadTreeNode(QuadTreeNode *&qtn);
QuadTreeNode(int posx, int posy, int wd, int ht);
bool VarianceCalculate(int Tolerance);
bool RandomPuzzyTag();
int AverageR();
int AverageG();
int AverageB();
void Fuzzify();
};
void QuadTreeNode::Fuzzify()
{
int average_r = AverageR();
int average_g = AverageG();
int average_b = AverageB();
for (int i = 0; i < height; ++i)
for (int j = 0; j < width; ++j)
{
rgbs[i][j].r = average_r;
rgbs[i][j].g = average_g;
rgbs[i][j].b = average_b;
}
valid = true;
}
bool QuadTreeNode::RandomPuzzyTag()
{return depth>5&&((int)rand()%6==0);}
bool QuadTreeNode::VarianceCalculate(int Tolerance) // 是否要继续细化
{
long long sum = height * width;
if(sum<=0)return true;
long long v = 0, res;
long average_r = AverageR();
for (int i = 0; i < height; ++i)
for (int j = 0; j < width; ++j)
v += pow(abs((int)rgbs[i][j].r - average_r), 2);
// res = abs(v / sum);
res= v/sum;
if (res >= pow(Tolerance, 2)-pow((8-depth)>0?8-depth:0,3))
//if (res >= -1)
return false;
v = 0;
int average_g = AverageG();
for (int i = 0; i < height; ++i)
for (int j = 0; j < width; ++j)
v += pow(abs((int)rgbs[i][j].g - average_g), 2);
// res = abs(v / sum);
res= v/sum;
if (res >= pow(Tolerance, 2)-pow((8-depth)>0?8-depth:0,3))
return false;
v = 0;
int average_b = AverageB();
for (int i = 0; i < height; ++i)
for (int j = 0; j < width; ++j)
v += pow(abs((int)rgbs[i][j].b - average_b), 2);
// res = abs(v / sum);
res= v / sum;
if (res >= pow(Tolerance, 2)-pow((8-depth)>0?8-depth:0,3))
return false;
return true;
}
QuadTreeNode::QuadTreeNode(int posx, int posy, int ht, int wd) : q1(NULL), q2(NULL), q3(NULL), q4(NULL), valid(false)
{
position.x = posx;
position.y = posy;
depth=0;
width = wd;
height = ht;
rgbs = (color **)malloc(height * sizeof(color *));
for (int i = 0; i < height; i++)
rgbs[i] = (color *)malloc(width * sizeof(color));
}
QuadTreeNode::QuadTreeNode(color **r, int wd, int ht) : q1(NULL), q2(NULL), q3(NULL), q4(NULL), valid(false)
{
width = wd;
height = ht;
depth=0;
position.x = 0;
position.y = 0;
rgbs = (color **)malloc(height * sizeof(color *));
for (int i = 0; i < height; i++)
rgbs[i] = (color *)malloc(width * sizeof(color));
for (int i = 0; i < height; ++i)
for (int j = 0; j < width; ++j)
rgbs[i][j] = r[i][j];
}
int QuadTreeNode::AverageR()
{
int sumr = 0;
for (int i = 0; i < height; ++i)
for (int j = 0; j < width; ++j)
sumr += ((int)rgbs[i][j].r);
return sumr / height / width;
}
int QuadTreeNode::AverageG()
{
int sumg = 0;
for (int i = 0; i < height; ++i)
for (int j = 0; j < width; ++j)
sumg += ((int)rgbs[i][j].g);
return sumg / height / width;
}
int QuadTreeNode::AverageB()
{
int sumb = 0;
for (int i = 0; i < height; ++i)
for (int j = 0; j < width; ++j)
sumb += ((int)rgbs[i][j].b);
return sumb / height / width;
}
#endif
// QuadTreeNode::QuadTreeNode(QuadTreeNode *&qtn)
// {
// position = qtn->position;
// height = qtn->height;
// width = qtn->width;
// valid = qtn->valid;
// q1 = qtn->q1;
// q2 = qtn->q2;
// q3 = qtn->q3;
// q4 = qtn->q4;
// rgbs = (color **)malloc(height * sizeof(color *));
// for (int i = 0; i < height; i++)
// rgbs[i] = (color *)malloc(width * sizeof(color));
// for (int i = 0; i < height; ++i)
// for (int j = 0; j < width; ++j)
// rgbs[i][j] = qtn->rgbs[i][j];
// }
----------------------------------QuadTreeFunc.h------------------------------------
#ifndef __QUADTREEFUNC_H__
#define __QUADTREEFUNC_H__
#include "QuadTreeNode.h"
struct color;
void TreeToImage(QuadTreeNode *rt, color **img)
{
if (rt->valid)
{
for (int i = 0; i < rt->height; ++i)
for (int j = 0; j < rt->width; ++j)
{
img[i + rt->position.x][j + rt->position.y].r = rt->rgbs[i][j].r;
img[i + rt->position.x][j + rt->position.y].g = rt->rgbs[i][j].g;
img[i + rt->position.x][j + rt->position.y].b = rt->rgbs[i][j].b;
}
}
else
{
TreeToImage(rt->q1, img);
TreeToImage(rt->q2, img);
TreeToImage(rt->q3, img);
TreeToImage(rt->q4, img);
}
}
void TreeFuzzifyExtend(QuadTreeNode *rt, int Tolerance)
{
if (rt->VarianceCalculate(Tolerance) || rt->width < 10 || rt->height < 10)
// if(rt->RandomPuzzyTag() || rt->width < 10 || rt->height < 10)
{
rt->Fuzzify();
rt->valid = true;
return;
}
int midwidth = 0, midheight = 0;
rt->valid = false;
midwidth = rt->width / 2;
midheight = rt->height / 2;
// subTree1
rt->q1 = new QuadTreeNode(rt->position.x, rt->position.y, midheight, midwidth);
rt->q1->valid = false;
rt->q1->depth = rt->depth + 1;
for (int i = 0; i < midheight; ++i)
for (int j = 0; j < midwidth; ++j)
rt->q1->rgbs[i][j] = rt->rgbs[i][j];
// subTree2
rt->q2 = new QuadTreeNode(rt->position.x, rt->position.y + midwidth, midheight, rt->width - midwidth);
rt->q2->valid = false;
rt->q2->depth = rt->depth + 1;
for (int i = 0; i < midheight; ++i)
for (int j = 0; j < rt->width - midwidth; ++j)
rt->q2->rgbs[i][j] = rt->rgbs[i][midwidth + j];
// subTree3
rt->q3 = new QuadTreeNode(rt->position.x + midheight, rt->position.y, rt->height - midheight, midwidth);
rt->q3->valid = false;
rt->q3->depth = rt->depth + 1;
for (int i = 0; i < rt->height - midheight; ++i)
for (int j = 0; j < midwidth; ++j)
rt->q3->rgbs[i][j] = rt->rgbs[midheight + i][j];
// subTree4
rt->q4 = new QuadTreeNode(rt->position.x + midheight, rt->position.y + midwidth, rt->height - midheight, rt->width - midwidth);
rt->q4->valid = false;
rt->q4->depth = rt->depth + 1;
for (int i = 0; i < rt->height - midheight; ++i)
for (int j = 0; j < rt->width - midwidth; ++j)
rt->q4->rgbs[i][j] = rt->rgbs[midheight + i][midwidth + j];
// 递归扩展
TreeFuzzifyExtend(rt->q1, Tolerance);
TreeFuzzifyExtend(rt->q2, Tolerance);
TreeFuzzifyExtend(rt->q3, Tolerance);
TreeFuzzifyExtend(rt->q4, Tolerance);
}
#endif
----------------------------------QuadImg_todo.cpp--------------------------------
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include "QuadTreeNode.h"
// #include "QuadTree.h"
#include "QuadTreeFunc.h"
void printImage(char *fileName, int width, color **a);
// color color;
// To get ppm image from jpeg file, please visit https://convertio.co/jpg-ppm/
void readImage(int p, char *inFile, char *outFile) // Note that width == height here
{
FILE *f = fopen(inFile, "rb");
char u[3]; // placehoder
int width, height, max_value;
fscanf(f, "%s%d%d%d%c", u, &width, &height, &max_value, &u[0]);
int i;
color **colors, **img;
colors = (color **)malloc(height * sizeof(color *));
img = (color **)malloc(height * sizeof(color *));
for (i = 0; i < height; i++)
{
colors[i] = (color *)malloc(width * sizeof(color));
img[i] = (color *)malloc(width * sizeof(color));
}
for (i = 0; i < height; i++)
fread(colors[i], sizeof(color), width, f);
fclose(f);
//=============================================================
QuadTreeNode *rt = new QuadTreeNode(colors, width, height);
TreeFuzzifyExtend(rt,15);
TreeToImage(rt, img);
//=============================================================
printImage(outFile, width, img);
}
//注意!!!/
//因为题目所给的图像长宽相等,所以此处输出时,只传入了width参数,视作height=width!! //
//实际作为一个小型功能软件时需要进行修改! //
//事实上,readImage()函数本来也视作height=width,不过我已经修改过了,此处需自行修改 //
void printImage(char *fileName, int width, color **a) // Note that width == height here
{
FILE *f = fopen(fileName, "wb");
fprintf(f, "P6\n");
fprintf(f, "%d %d\n", width, width);
fprintf(f, "255\n");
int i;
for (i = 0; i < width; i++)
fwrite(a[i], sizeof(color), width, f);
fclose(f);
}
// int main(int argc, char **argv)
int main()
{
int tolerance = 0;
char inFile[100];
char outFile[100];
// if (argc > 1)
{
// tolerance = atoi(argv[1]);
// inFile = argv[2];
// outFile = argv[3];
strcpy(inFile,"D:\\DataStructuresAndAlgorithms\\homework\\Experiment3-Quadtree-adaptive-fuzzy\\a.ppm");
strcpy(outFile,"D:\\DataStructuresAndAlgorithms\\homework\\Experiment3-Quadtree-adaptive-fuzzy\\result.ppm");
readImage(tolerance, inFile, outFile);
}
return 0;
}
到了这里,关于【数据结构与算法】图像——四叉树自适应模糊(ppm图像为例)的文章就介绍完了。如果您还想了解更多内容,请在右上角搜索TOY模板网以前的文章或继续浏览下面的相关文章,希望大家以后多多支持TOY模板网!