Browse Source

Adds hilbert/wavelet thingies.

master
Joshua Moerman 8 years ago
parent
commit
0b6f80473a
  1. 9
      CMakeLists.txt
  2. 109
      wavelet/jcmp.cpp
  3. 5
      wavelet/jcmp_quantization.hpp
  4. 100
      wavelet/remap.hpp
  5. 66
      wavelet/remap_test.cpp

9
CMakeLists.txt

@ -1,14 +1,16 @@
cmake_minimum_required (VERSION 2.6)
project(compress)
include_directories(SYSTEM "${PROJECT_SOURCE_DIR}/include/")
add_definitions( -std=c++0x )
set(CMAKE_CXX_FLAGS ${CMAKE_CXX_FLAGS} "-std=c++11")
option(MultiCoreBSP "Building with Multi Core BSP" OFF)
option(MultiCoreBSP "Building with Multi Core BSP" ON)
if(MultiCoreBSP)
# setup for my mac
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
set(libs mcbsp1.1.0 m pthread)
set(libs mcbsp m pthread)
add_definitions( -DUSE_MCBSP )
include_directories(SYSTEM "${PROJECT_SOURCE_DIR}/contrib/mcbsp")
else()
# setup for cartesius
set(libs bsponmpi m)
@ -19,3 +21,4 @@ include_directories(SYSTEM ${Boost_INCLUDE_DIRS})
set(libs ${libs} ${Boost_LIBRARIES})
add_subdirectory("wavelet")
add_subdirectory("contrib/mcbsp")

109
wavelet/jcmp.cpp

@ -1,21 +1,38 @@
#include <includes.hpp>
#include <utilities.hpp>
#include "jcmp.hpp"
#include "wavelet.hpp"
#include "remap.hpp"
#include <boost/filesystem.hpp>
#include <boost/program_options.hpp>
#include <png.hpp>
#include "jcmp.hpp"
#include "wavelet.hpp"
#include <vector>
#include <iostream>
using namespace wvlt;
using namespace jcmp;
// note: we take a copy, because we will modify it in place
static image compress(std::vector<double> img, uint width, double threshold, unsigned int& nzeros){
uint height = img.size() / width;
static image compress(std::vector<double> const & img0, uint16_t width, double threshold, unsigned int& nzeros){
uint16_t height = img0.size() / width;
assert(is_pow_of_two(width));
assert(is_pow_of_two(height));
assert(width == height);
std::vector<double> img(img0.size(), 0);
for(uint16_t y = 0; y < height; y++){
for(uint16_t x = 0; x < width; x++) {
img[remap::to_hilbert(width, x, y)] = img0[x + y*width];
}
}
wavelet(&img[0], width * height, 1);
#if 0
// wavelet transform in x-direction
for(unsigned int i = 0; i < height; ++i){
wavelet(&img[i*width], width, 1);
@ -25,6 +42,7 @@ static image compress(std::vector<double> img, uint width, double threshold, uns
for(unsigned int i = 0; i < width; ++i){
wavelet(&img[i], height, width);
}
#endif
double min_abs = 10000.0;
double max_abs = 0.0;
@ -43,8 +61,8 @@ static image compress(std::vector<double> img, uint width, double threshold, uns
// save the principal coefficients
std::vector<jcmp::coefficient> v;
for(unsigned int y = 0; y < height; ++y){
for(unsigned int x = 0; x < width; ++x){
for(uint16_t y = 0; y < height; ++y){
for(uint16_t x = 0; x < width; ++x){
auto&& el = img[x + width*y];
if(el != 0) v.push_back({q.forwards(el), uint(x), uint(y)});
}
@ -55,8 +73,8 @@ static image compress(std::vector<double> img, uint width, double threshold, uns
}
static std::vector<double> decompress(image in){
auto width = in.header.width;
auto height = in.header.height;
const auto width = in.header.width;
const auto height = in.header.height;
assert(is_pow_of_two(width));
assert(is_pow_of_two(height));
@ -66,12 +84,13 @@ static std::vector<double> decompress(image in){
// read in coefficient on coordinates
for(auto it = in.data.begin(); it != in.data.end(); ++it){
auto&& x = *it;
image[x.x + width*x.y] = q.backwards(x.c);
auto&& coef = *it;
image[coef.x + width*coef.y] = q.backwards(coef.c);
}
in.clear();
#if 0
// inverse wavelet transform in y-direction
for(unsigned int i = 0; i < width; ++i){
unwavelet(&image[i], height, width);
@ -81,18 +100,70 @@ static std::vector<double> decompress(image in){
for(unsigned int i = 0; i < height; ++i){
unwavelet(&image[i*width], width, 1);
}
#endif
unwavelet(&image[0], width*height, 1);
std::vector<double> image2(width * height, 0.0);
for(uint32_t d = 0; d < width * height; ++d) {
const auto p = remap::from_hilbert(width, d);
image2[p.first + width * p.second] = image[d];
}
return image;
return image2;
}
int main(){
int main(int argc, char* argv[]){
using namespace std;
namespace fs = boost::filesystem;
fs::path directory("images");
fs::directory_iterator eod;
for(fs::directory_iterator it(directory); it != eod; ++it){
auto && path = it->path();
if(path.extension() != ".png") continue;
struct {
vector<fs::path> filenames = {};
double threshold = 5.0;
} options;
namespace po = boost::program_options;
// Describe program options
po::options_description opts;
opts.add_options()
("threshold", po::value<double>(), "")
("help", po::bool_switch(), "show this help")
("file", po::value< vector<fs::path> >(), "input file(s)");
// and inputs
po::positional_options_description p;
p.add("file", -1);
// Parse and set options
try {
po::variables_map vm;
po::store(po::command_line_parser(argc, argv).
options(opts).positional(p).run(), vm);
po::notify(vm);
if(vm["help"].as<bool>()){
std::cout << "jcmp" << std::endl;
std::cout << opts << std::endl;
return 0;
}
if(vm.count("file")){
options.filenames = vm["file"].as< vector<fs::path> >();
}
if(vm.count("threshold")) {
options.threshold = vm["threshold"].as<double>();
}
} catch(std::exception& e){
std::cout << colors::red("ERROR: ") << e.what() << std::endl;
std::cout << opts << std::endl;
return 1;
}
for(const auto & path : options.filenames) {
if(path.extension() != ".png") {
cerr << "WARNING: can only read .png, but " << path << " is not" << endl;
}
// open file
std::string filename = path.string();
@ -109,7 +180,7 @@ int main(){
// compress and decompress to see how we lost information
unsigned int nzeros = 0;
auto compressed_vec = decompress(compress(image_vec, width, 5.0, nzeros));
auto compressed_vec = decompress(compress(image_vec, width, options.threshold, nzeros));
// output some information
std::cout << field("raw") << width*height << std::endl;
@ -117,7 +188,7 @@ int main(){
std::cout << field("nz") << nzeros << std::endl;
// save to output file
std::string cfilename = "compressed/" + std::to_string(nzeros) + path.filename().string();
std::string cfilename = "hil_compressed_" + std::to_string(nzeros) + path.filename().string();
png::gray_ostream compressed_image(width, height, cfilename);
for(unsigned int i = 0; i < compressed_vec.size(); ++i) compressed_image << compressed_vec[i];
}

5
wavelet/jcmp_quantization.hpp

@ -50,6 +50,7 @@ namespace jcmp {
enum class type {
linear,
logarithmic,
quadric_root,
square_root
};
@ -71,6 +72,10 @@ namespace jcmp {
b.f = &log;
b.f_inv = &exp;
break;
case type::quadric_root:
b.f = [](double x){ return 1 - 1/(0.25*x*x + 1); };
b.f_inv = [](double x) { return std::sqrt(1/(1 - x)-1); };
break;
case type::square_root:
b.f = &sqrt;
b.f_inv = &sqr;

100
wavelet/remap.hpp

@ -0,0 +1,100 @@
#pragma once
#include <climits>
#include <cstdint>
#include <utility>
namespace remap {
// there are some more efficient ways in
// http://graphics.stanford.edu/~seander/bithacks.html#InterleaveTableObvious
// but I was too lazy to understand them
// some templates dealing with sizes
// clang-format off
template <typename T> struct unsigned_size {};
template <> struct unsigned_size<uint8_t> { using twice = uint16_t; /* 4 bits */ };
template <> struct unsigned_size<uint16_t> { using twice = uint32_t; using half = uint8_t; };
template <> struct unsigned_size<uint32_t> { using twice = uint64_t; using half = uint16_t; };
template <> struct unsigned_size<uint64_t> { /* 128 bits */ using half = uint32_t; };
// clang-format on
// shortcuts
template <typename T>
using twice = typename unsigned_size<T>::twice;
template <typename T>
using half = typename unsigned_size<T>::half;
// converts x,y coordinates to z order
template <typename T>
twice<T> to_z_order(const T xin, const T yin) {
const twice<T> x = xin;
const twice<T> y = yin;
twice<T> z = 0;
for (int i = 0; i < sizeof(T) * CHAR_BIT; i++) {
z |= (x & (1U << i)) << i | (y & (1U << i)) << (i + 1);
}
return z;
}
// converts z to x,y
template <typename T>
std::pair<half<T>, half<T>> from_z_order(const T z) {
half<T> x = 0;
half<T> y = 0;
for (int i = 0; i < sizeof(T) * CHAR_BIT; i += 2) {
x |= (z & (1U << i)) >> (i / 2);
y |= (z & (1U << (i + 1))) >> (i / 2 + 1);
}
return {x, y};
}
// from wikipedia
// rotate/flip a quadrant appropriately
template <typename T>
void hilbert_rot(T n, T * x, T * y, T rx, T ry) {
if (ry == 0) {
if (rx == 1) {
*x = n - 1 - *x;
*y = n - 1 - *y;
}
// Swap x and y
int t = *x;
*x = *y;
*y = t;
}
}
template <typename T>
twice<T> to_hilbert(twice<T> n, T xin, T yin) {
twice<T> x = xin;
twice<T> y = yin;
twice<T> d = 0;
for (twice<T> s = n / 2; s > 0; s /= 2) {
twice<T> rx = (x & s) > 0;
twice<T> ry = (y & s) > 0;
d += s * s * ((3 * rx) ^ ry);
hilbert_rot(s, &x, &y, rx, ry);
}
return d;
}
// convert d to (x,y)
template <typename T>
std::pair<half<T>, half<T>> from_hilbert(T n, T d) {
T t = d;
T x = 0;
T y = 0;
for (T s = 1; s < n; s *= 2) {
T rx = 1 & (t / 2);
T ry = 1 & (t ^ rx);
hilbert_rot(s, &x, &y, rx, ry);
x += s * rx;
y += s * ry;
t /= 4;
}
return {half<T>(x), half<T>(y)};
}
}

66
wavelet/remap_test.cpp

@ -0,0 +1,66 @@
#include "remap.hpp"
#include <iostream>
#include <limits>
using namespace std;
template <typename T>
static void test_z_order(uint64_t xmax, uint64_t ymax){
uint64_t c = 10;
for (uint64_t x2 = 0; x2 <= xmax; ++x2) {
for (uint64_t y2 = 0; y2 <= ymax; ++y2) {
const T x = x2;
const T y = y2;
const auto z = remap::to_z_order(x, y);
const auto p = remap::from_z_order(z);
if (x != p.first || y != p.second) {
cout << +x << ' ' << +y << " -> " << z << " -> " << +p.first << ' ' << +p.second << '\n';
c--;
if(!c) return;
}
}
}
}
template <typename T>
static void test_z_order_max(){
return test_z_order<T>(std::numeric_limits<T>::max(), std::numeric_limits<T>::max());
}
template <typename T>
static void test_hilbert(uint64_t xmax){
uint64_t c = 10;
for (uint64_t x2 = 0; x2 <= xmax; ++x2) {
for (uint64_t y2 = 0; y2 <= xmax; ++y2) {
const T x = x2;
const T y = y2;
const remap::twice<T> xmax2 = xmax+1;
const auto z = remap::to_hilbert(xmax2, x, y);
const auto p = remap::from_hilbert(xmax2, z);
if (x != p.first || y != p.second) {
cout << +x << ' ' << +y << " -> " << z << " -> " << +p.first << ' ' << +p.second << '\n';
c--;
if(!c) return;
}
}
}
}
template <typename T>
static void test_hilbert_max(){
return test_hilbert<T>(std::numeric_limits<T>::max());
}
int main(int argc, char * argv[]) {
//cout << "Z order: 8" << endl;
//test_z_order_max<uint8_t>();
cout << "Hilbert: 8" << endl;
test_hilbert_max<uint8_t>();
//cout << "Z order: 16" << endl;
//test_z_order_max<uint16_t>();
cout << "Hilbert: 16" << endl;
test_hilbert_max<uint16_t>();
}