/** * Image compression using NNs * $Id: ImageCompress.h,v 1.1 2004/06/16 10:53:30 opx Exp $ * @author OP */#ifndef IMAGE_COMPRESS_H#define IMAGE_COMPRESS_H#include "annie/defines.h"#include "annie/MultiLayerNetwork.h"#include <vector>namespace annie	{class TrainingSet;class Image;/** * (usually lossy) image compressor. *//*abstract */class ImageCompressor	{  public:	virtual ~ImageCompressor() {}	//be happy, my dear compiler :)	class PartialListener	{	  public:		virtual void showProgress(const Image &partialResult) = 0;	};	virtual Image *compress(const Image &i) = 0;	/** 	 * Use this if you want to display the partial results.	 * Note that it's up to implementation if and when it will call listener.	 */	void setPartialListener(PartialListener *pl)	{   _partialListener = pl; }  protected:	void showPartialProgress(const Image &partialResult)	{ if( _partialListener)  _partialListener->showProgress(partialResult); }	ImageCompressor() :  _partialListener(NULL) {}  private:	PartialListener *_partialListener;};///container for imagestypedef std::vector<annie::Image *> Images;/** * Compress image by dividing it to squares. * Implementations will typically choose a number of distinct squares *//*abstract */class TileCompressor : public ImageCompressor	{  public:	TileCompressor(uint partSize) : _partSize(partSize) {}	uint getPartSize()	const {	return _partSize; }	virtual Image *compress(const Image &i);	virtual Image *compressTiles(const Images &images) = 0; 	/**	 * Cut the image into the parts	 */	Images cut(const Image &source) const;  protected:	uint _partSize;	//<some implementations may allow to change it during lifetime	uint _px, _py;	//< valid only during compress - count of cols and rows};/**  * the input image should be grayed because we only take the first color component into account * supports partial progress displaying */ class MLPCompressor : public TileCompressor, public ValueUpdateListener		///yeah, yeah - ValueUpdateListener	should have been inner class{  public:	enum { RESHOW = 20 };	//redraw period	/**	 * @param hidden? number of neurons in the respective layer - if 0, no such layer	 */	MLPCompressor(uint partSize, uint hidden1, uint hidden2, PublicValues &pv = defaultControl);	virtual Image *compressTiles(const Images &images);	virtual void valueChanged(const Value &val);		///direct manipulation with the net - not needed for standard opraration. be careful!	MultiLayerNetwork &getNet() { return _net; }	///if you want the examples used inside ..	void fillExamples(const Images &images, TrainingSet &dest);	//TS is not a result because pass by value of TS is slowwwww  protected:	MultiLayerNetwork _net;	std::vector<Vector> vectored;	//< vectored tiles	//construct result using vectored and the net	Image *getResult();	Creal _reshow;	//re-show period};				// utility functions for creating images	Images makeSubimages(const annie::Image &i, unsigned xSize, unsigned ySize);	/// make vector from each tile. Only the 1st byte of each color component is taken into account (to reduce dimensionalyty for MLP)	/// retuens the outExamples and the same stored as identity in TrainingSet	TrainingSet *makeSubImageExamples(const Images &outImages, std::vector<Vector> &outExamples);	/**	 * generic ImageCompressor utilization	 * uses ["input"] as the input file	 * @param gray whether to gray the input image before processing	 */	void runCompressor(ImageCompressor &ic, const char *image, bool graay=false);}	//annie#endif