Fixed downscaling metric

lib/fingerprints/downscale.cpp

@@ -1,11 +1,39 @@
 #include "downscale.hpp"
 
 #include <image_io.hpp>
-#include <algorithm>
 
-downscale downscale::pre_calculate(const av::frame& frame){
+extern "C" {
-	// ffmpeg doesnt let us downscale all the way to 5 at once :(
+#include <libavutil/frame.h>
-	auto const image = to_raw_rgb_image(frame, 5, 5);
+#include <libswscale/swscale.h>
+}
+
+#include <algorithm>
+#include <iostream>
+
+static raw_rgb_image downscale_step(AVFrame const * frame, int factor) {
+	raw_rgb_image image(frame->width / factor, frame->height / factor);
+
+	auto context = sws_getContext(frame->width, frame->height, static_cast<AVPixelFormat>(frame->format), image.width(), image.height(), image.format(), 0, nullptr, nullptr, nullptr);
+	if(!context) throw std::runtime_error("boem sws context");
+	sws_scale (context, {frame->data}, {frame->linesize}, 0, frame->height, {image.frame->data}, {image.frame->linesize});
+	sws_freeContext(context);
+
+	return image;
+}
+
+static raw_rgb_image downscale_to(const av::frame &frame, int w, int h){
+	// ffmpeg doesnt let us downscale all the way to 5 at once :(, so we do a loop
+	raw_rgb_image image;
+	auto new_frame = frame.get();
+	while(new_frame->width > 8*w && new_frame->height > 8*h){
+		image = downscale_step(new_frame, 4);
+		new_frame = image.frame.get();
+	}
+	return to_raw_rgb_image(image.frame, w, h);
+}
+
+downscale downscale::pre_calculate(av::frame const & frame){
+	auto const image = downscale_to(crop_to_square(frame), 5, 5);
 
 	downscale ret;
 	ret.data.assign(image.data.size(), 0);
@@ -19,7 +47,7 @@ downscale downscale::calculate(const av::frame& frame){
 	return pre_calculate(frame);
 }
 
-double square(double x){
+static double square(double x){
 	return x*x;
 }
 

lib/image_io.cpp

@@ -15,6 +15,11 @@ extern "C" {
 #include <cassert>
 
 
+raw_rgb_image::raw_rgb_image()
+: data()
+, frame(nullptr, [](auto x){ av_frame_free(&x); })
+{}
+
 raw_rgb_image::raw_rgb_image(int W, int H)
 : data(make_u(avpicture_get_size(AV_PIX_FMT_RGB24, W, H)))
 , frame(av::frame_alloc())

lib/image_io.hpp

@@ -13,6 +13,7 @@ struct raw_rgb_image {
 	std::vector<uint8_t, av::allocator<uint8_t>> data;
 	av::frame frame;
 
+	raw_rgb_image();
 	raw_rgb_image(int W, int H);
 
 	int width() const;

src/main.cpp

@@ -28,10 +28,10 @@ using namespace std;
 namespace fs = boost::filesystem;
 namespace ar = boost::archive;
 
-using Database = image_database<rgb_wavelet_coefficients>;
+using Database = image_database<downscale>;
 using Mozaic = map<pair<int, int>, string>;
 
-Database read_database(string const & database_directory){
+static Database read_database(string const & database_directory){
 	Database db;
 	auto const database_file = database_directory + ".db";
 
@@ -60,10 +60,10 @@ Database read_database(string const & database_directory){
 	return db;
 }
 
-Mozaic create_mozaic(Database const & db, string const & filename, int h_tiles, int v_tiles){
+static Mozaic create_mozaic(Database const & db, string const & filename, int h_tiles, int v_tiles){
 	Mozaic mozaic;
 
-	apply_to_tiles("image.jpg", h_tiles, v_tiles, [&](int c, int r, av::frame const & frame){
+	apply_to_tiles(filename, h_tiles, v_tiles, [&](int c, int r, av::frame const & frame){
 		auto const index = db.nearest_image(frame);
 		cout << colors::red("tile ") << c << ", " << r << ": " << db.filename(index) << endl;
 		mozaic[make_pair(c, r)] = db.filename(index);
@@ -72,7 +72,7 @@ Mozaic create_mozaic(Database const & db, string const & filename, int h_tiles,
 	return mozaic;
 }
 
-void save_mozaic(Mozaic const & mozaic, string filename, int h_tiles, int v_tiles){
+static void save_mozaic(Mozaic const & mozaic, string filename, int h_tiles, int v_tiles){
 	auto const pix_fmt = AV_PIX_FMT_YUVJ444P;
 	auto const codec_id= AV_CODEC_ID_MJPEG;