Coursera 算法二 week2 Seam Carving
这周作业设计到的算法是有向无环图的最短路径算法,只需要按照顶点的拓扑顺序去放松顶点即可。而在这个题目中拓扑顺序就是按照行的顺序或列的顺序。
用到的数据结构为一个二维数组picture同来存储每个像素的颜色,一个二维数组energy用来存储每个像素的能量。开始我是用一个Picture类的对象来存储图像,但是在讨论区里发现用二维数组存储图像,可以占用更小的存储,且计算能量、removeseam时更快更方便。
在检验各像素能量时发现计算结果不正确,后来发现是运算符优先级的问题,((rgbLeft >> 16) & 0xFF) - ((rgbRight >> 16) & 0xFF),即‘ - ’的优先级大于‘ & ’的优先级,因此需要加括号。
在Checklist的Possible Progress Steps中发现计算seam以及removeseam时可以只写Horizontal和Vertical中的一个,然后另一个用矩阵转置的方法来完成。
第一次提交时memory没有通过,原因是把二维数组distTo和二维数组edgeTo放到了成员变量里,后来把这两个数组放到局部变量,就通过了memory测试。
import edu.princeton.cs.algs4.Picture;
public class SeamCarver {
private int[][] picture;
private double[][] energy;
private int width;
private int height;
public SeamCarver(Picture picture) // create a seam carver object based on the given picture
{
if (picture == null)
throw new IllegalArgumentException();
width = picture.width();
height = picture.height();
energy = new double[width][height];
this.picture = new int[width][height];
for (int i = 0; i < width(); i++)
{
for (int j = 0; j < height(); j++)
this.picture[i][j] = picture.getRGB(i, j);
}
for (int i = 0; i < width(); i++)
{
for (int j = 0; j < height(); j++)
energy[i][j] = computeEnergy(i, j);
}
}
private double computeEnergy(int x, int y)
{
if (x == 0 || x == width() - 1 || y == 0 || y == height() - 1)
return 1000.0;
int rgbUp = picture[x][y - 1];
int rgbDown = picture[x][y + 1];
int rgbLeft = picture[x - 1][y];
int rgbRight = picture[x + 1][y];
double rx = Math.pow(((rgbLeft >> 16) & 0xFF) - ((rgbRight >> 16) & 0xFF), 2);
double gx = Math.pow(((rgbLeft >> 8) & 0xFF) - ((rgbRight >> 8) & 0xFF), 2);
double bx = Math.pow(((rgbLeft >> 0) & 0xFF) - ((rgbRight >> 0) & 0xFF), 2);
double ry = Math.pow(((rgbUp >> 16) & 0xFF) - ((rgbDown >> 16) & 0xFF), 2);
double gy = Math.pow(((rgbUp >> 8) & 0xFF) - ((rgbDown >> 8) & 0xFF), 2);
double by = Math.pow(((rgbUp >> 0) & 0xFF) - ((rgbDown >> 0) & 0xFF), 2);
return Math.sqrt(rx + gx + bx + ry + gy + by);
}
public Picture picture() // current picture
{
Picture pic = new Picture(width, height);
for (int i = 0; i < width; i++)
for (int j = 0; j < height; j++)
pic.setRGB(i, j, picture[i][j]);
return pic;
}
public int width() // width of current picture
{
return width;
}
public int height() // height of current picture
{
return height;
}
public double energy(int x, int y) // energy of pixel at column x and row y
{
if (x < 0 || x > width - 1 || y < 0 || y > height - 1)
throw new IllegalArgumentException();
return energy[x][y];
}
private void relaxvertical(double[][] distTo, int[][] edgeTo, int x, int y)
{
if (distTo[x][y + 1] > distTo[x][y] + energy[x][y + 1])
{
distTo[x][y + 1] = distTo[x][y] + energy[x][y + 1];
edgeTo[x][y + 1] = x;
}
if (x > 0 && distTo[x - 1][y + 1] > distTo[x][y] + energy[x - 1][y + 1])
{
distTo[x - 1][y + 1] = distTo[x][y] + energy[x - 1][y + 1];
edgeTo[x - 1][y + 1] = x;
}
if (x < width() - 1 && distTo[x + 1][y + 1] > distTo[x][y] + energy[x + 1][y + 1])
{
distTo[x + 1][y + 1] = distTo[x][y] + energy[x + 1][y + 1];
edgeTo[x + 1][y + 1] = x;
}
}
private void transpose()
{
int temp = width;
width = height;
height = temp;
double[][] energy2 = new double[width][height];
int[][] picture2 = new int[width][height];
for (int i = 0; i < width; i++)
{
for (int j = 0; j < height; j++)
{
energy2[i][j] = energy[j][i];
picture2[i][j] = picture[j][i];
}
}
energy = energy2;
picture = picture2;
}
public int[] findHorizontalSeam() // sequence of indices for horizontal seam
{
transpose();
int[] array = findVerticalSeam();
transpose();
return array;
}
public int[] findVerticalSeam() // sequence of indices for vertical seam
{
int[] seam = new int[height];
double[][] distTo = new double[width][height];
int[][] edgeTo = new int[width][height];
for (int i = 0; i < width; i++)
{
for (int j = 0; j < height; j++)
{
if (j == 0) distTo[i][j] = energy[i][j];
else distTo[i][j] = Double.POSITIVE_INFINITY;
}
}
for (int j = 0; j < height - 1; j++)
{
for (int i = 0; i < width; i++)
{
relaxvertical(distTo, edgeTo, i, j);
}
}
double min = Double.MAX_VALUE;
int minIndex = 0;
for (int i = 0; i < width; i++)
{
if (distTo[i][height - 1] < min)
{
min = distTo[i][height - 1];
minIndex = i;
}
}
seam[height - 1] = minIndex;
for (int j = height - 2; j >= 0; j--)
{
seam[j] = edgeTo[seam[j + 1]][j + 1];
}
return seam;
}
public void removeHorizontalSeam(int[] seam) // remove horizontal seam from current picture
{
checkSeam(seam);
int min = Integer.MAX_VALUE;
int max = 0;
for (int i = 0; i < width; i++)
{
if (seam[i] > max) max = seam[i];
if (seam[i] < min) min = seam[i];
for (int j = seam[i]; j < height - 1; j++)
{
picture[i][j] = picture[i][j + 1];
}
}
height--;
if (min > 0) min--;
if (max > height - 1) max = height - 1;
for (int i = 0; i < width; i++)
{
for (int j = min; j <= max; j++)
energy[i][j] = computeEnergy(i, j);
for (int j = max + 1; j < height - 1; j++)
energy[i][j] = energy[i][j + 1];
}
}
private void checkSeam(int[] seam)
{
if (height <= 1 || seam == null || seam.length != width)
throw new IllegalArgumentException();
for (int i = 0; i < width; i++)
{
if (seam[i] < 0 || seam[i] > height - 1)
throw new IllegalArgumentException();
if (i > 0 && Math.abs(seam[i] - seam[i - 1]) > 1)
throw new IllegalArgumentException();
}
}
public void removeVerticalSeam(int[] seam) // remove vertical seam from current picture
{
transpose();
removeHorizontalSeam(seam);
transpose();
}
}
Coursera 算法二 week2 Seam Carving的更多相关文章
- Coursera 算法二 week 5 BurrowsWheeler
本打算周末完成这次作业,但没想到遇到了hard deadline,刚开始看不懂题意,后来发现算法4书上有个类似的问题,才理解了题意.最后晚上加班,上课加班,还好在11:35也就是课程结束前25分钟完成 ...
- Coursera 算法二 week 3 Baseball Elimination
这周的作业不需要自己写算法,只需要调用库函数就行,但是有些难以理解,因此用了不少时间. import edu.princeton.cs.algs4.FlowEdge; import edu.princ ...
- Coursera 算法二 week 4 Boggle
这次的作业主要用到了单词查找树和深度优先搜索. 1.在深度优先搜索中,在当前层的递归调用前,将marked数组标记为true.当递归调用返回到当前层时,应将marked数组标记为false.这样既可以 ...
- coursera 算法二 week 1 wordnet
这周的作业可谓是一波三折,但是收获了不少,熟悉了广度优先搜索还有符号图的建立.此外还知道了Integer.MAX_VALUE. SAP: 求v和w的大概思路是对v和w分别广度优先搜索,然后遍历图中每一 ...
- Programming Assignment 2: Seam Carving
编程作业二 作业链接:Seam Carving & Checklist 我的代码:SeamCarver.java 问题简介 接缝裁剪(Seam carving),是一个可以针对照片内容做正确缩 ...
- Seam carving 学习笔记
今天首次接触了图像编辑中的seam carving知识,感觉挺神奇的.虽然我自己可能理解的不是很深刻,但是记录下来,总是好的. seam carving直接翻译过来是“线裁剪”的意思.它的主要用途是对 ...
- HDU5092——Seam Carving(动态规划+回溯)(2014上海邀请赛重现)
Seam Carving DescriptionFish likes to take photo with his friends. Several days ago, he found that s ...
- 递推DP HDOJ 5092 Seam Carving
题目传送门 /* 题意:从上到下,找最短路径,并输出路径 DP:类似数塔问题,上一行的三个方向更新dp,路径输出是关键 */ #include <cstdio> #include < ...
- TensorFlow 入门之手写识别(MNIST) softmax算法 二
TensorFlow 入门之手写识别(MNIST) softmax算法 二 MNIST Fly softmax回归 softmax回归算法 TensorFlow实现softmax softmax回归算 ...
随机推荐
- C# 中的EventHandler实例详解-转
//这里定义了一个水箱类 public class 水箱 { //这是水箱的放水操作 public void 放水() { } //这是水箱的属性 public double 体积; ...
- C# Object reference not set to an instance of an object.
一.问题 Object reference not set to an instance of an object. (你调用的对象是空的) 二.解决问题 在使用 c# 的查询时,先筛选后在关联其他表 ...
- Python数据聚合和分组运算(2)-Data Aggregation
在上一篇博客里我们讲解了在python里运用pandas对数据进行分组,这篇博客将接着讲解对分组后的数据进行聚合. 1.python 中经过优化的groupy方法 先读入本文要使用的数据集tips. ...
- Bloomberg 专线配置问题
其实很简单,满足bbg的要求就可以了, 配置如下几个网段的路由: 208.134.161.0 using the subnet mask of 255.255.255.0 205.183.246.0 ...
- c# Relative Path convert to Absolute Path
Reference: http://stackoverflow.com/questions/4796254/relative-path-to-absolute-path-in-c http://sta ...
- MCP|XHK|High-density peptide arrays help to identify linear immunogenic B cell epitopes in individuals naturally exposed to malaria infection(高密度肽段阵列有助于在自然暴露于疟疾感染的个体中识别线性免疫原性B细胞表位)
文献名:High-density peptide arrays help to identify linear immunogenic B cell epitopes in individuals n ...
- 使用shell脚本分析Nagios的status.dat文件
前言 Nagios的安装和配置以及批量添加监控服务器在我前面的文章中已经讲的很详细了. 我们知道,Nagios的网页控制页面(一般为http://nagio.domain.com/nagios)里可以 ...
- EOS区块同步源码分析之见证者
调了两天,终于把EOS的出块和DPOS共识机制的代码看了个大概,下面做个小总结.因为这一块是EOS比较核心的代码块,稍微复杂些,如果有理解错误之处,请大家见识并帮我指出来,我也会有可能随时改写这里的内 ...
- swarm
https://blog.51cto.com/lookingdream/2060292 一.规划 1.swarm01作为manager节点,swarm02和swarm03作为worker节点. # c ...
- C# ThreadLocal
ThreadLocal的主要作用是让各个线程维持自己的变量. .NET 4.0在线程方面加入了很多东西,其中就包括ThreadLocal<T>类型,他的出现更大的简化了TLS的操作.Thr ...