📄 grayf.hpp
字号:
/*************************************************************************
This software module was originally developed by
Ming-Chieh Lee (mingcl@microsoft.com), Microsoft Corporation
Wei-ge Chen (wchen@microsoft.com), Microsoft Corporation
Bruce Lin (blin@microsoft.com), Microsoft Corporation
Chuang Gu (chuanggu@microsoft.com), Microsoft Corporation
(date: March, 1996)
in the course of development of the MPEG-4 Video (ISO/IEC 14496-2).
This software module is an implementation of a part of one or more MPEG-4 Video tools
as specified by the MPEG-4 Video.
ISO/IEC gives users of the MPEG-4 Video free license to this software module or modifications
thereof for use in hardware or software products claiming conformance to the MPEG-4 Video.
Those intending to use this software module in hardware or software products are advised that its use may infringe existing patents.
The original developer of this software module and his/her company,
the subsequent editors and their companies,
and ISO/IEC have no liability for use of this software module or modifications thereof in an implementation.
Copyright is not released for non MPEG-4 Video conforming products.
Microsoft retains full right to use the code for his/her own purpose,
assign or donate the code to a third party and to inhibit third parties from using the code for non <MPEG standard> conforming products.
This copyright notice must be included in all copies or derivative works.
Copyright (c) 1996, 1997.
Module Name:
grayf.hpp
Abstract:
Float image class for gray (one-plane) pictures
Revision History:
*************************************************************************/
#ifndef __GRAYF_HPP_
#define __GRAYF_HPP_
typedef Float PixelF;
#define MAX_NUM_CONTOUR 200
Class CVideoObjectPlane;
Class CFloatImage
{
public:
// Constructors
~CFloatImage ();
CFloatImage (const CFloatImage& fi, const CRct& r = CRct ()); // crop from a floatimage
CFloatImage (const CIntImage& fi, const CRct& r = CRct ()); // crop from an int image
CFloatImage (const CVideoObjectPlane& vop, RGBA comp, const CRct& r = CRct ());
CFloatImage (const CRct& r = CRct (), PixelF px = 0.); // floatimage with same pixel value everywhere
CFloatImage (const CPolygonI& plg, const CRct& rct); // fill in the polygon (interior boundary)
CFloatImage (
const Char* pchFileName, // MPEG4 file format
UInt ifr, // frame number
const CRct& rct, // final rect
UInt nszHeader = 0 // number of header to be skipped
);
CFloatImage (const Char* vdlFileName); // read from a VM file.
// Attributes
Bool valid () const {return this != 0;}
const CRct& where () const {return m_rc;}
CRct boundingBox (const PixelF pxlfOutsideColor = (PixelF) transpValue) const;
PixelF pixel (CoordI x, CoordI y) const {return m_ppxlf [m_rc.offset (x, y)];}
PixelF pixel (const CSite& st) const {return pixel (st.x, st.y);}
const PixelF* pixels () const {return (this == NULL) ? NULL : m_ppxlf;} // return pointer
const PixelF* pixels (CoordI x, CoordI y) const {return m_ppxlf + m_rc.offset (x, y);}
const PixelF* pixels (const CSite& st) const {return pixels (st.x, st.y);}
// Resultants
own CFloatImage* decimate (UInt rateX, UInt rateY) const; // decimate the vframe by rateX and rateY.
own CFloatImage* decimateBinaryShape (UInt rateX, UInt rateY) const; // decimate by rateX and rateY in a conservative way
own CFloatImage* zoomup (UInt rateX, UInt rateY) const;
own CFloatImage* expand (UInt rateX, UInt rateY) const; // expand by putting zeros in between
own CFloatImage* biInterpolate () const; // bilinearly interpolate
own CFloatImage* downsampleForSpatialScalability () const;
own CFloatImage* upsampleForSpatialScalability () const;
own CFloatImage* biInterpolate (UInt accuracy) const; // bilinearly interpolate using 2/4/8/16 quantization
own CFloatImage* transpose () const; // transpose the floatimge
own CFloatImage* complement () const; // complement a binarized fi
own CFloatImage* warp (const CAffine2D& aff) const; // affine warp
own CFloatImage* warp (const CPerspective2D& persp) const; // perspective warp
own CFloatImage* warp (const CPerspective2D& persp, const CRct& rctWarp) const; // perspective warp to rctWarp
own CFloatImage* warp (const CPerspective2D& persp, const CRct& rctWarp, const UInt accuracy) const; // perspective warp to rctWarp
own CFloatImage* smooth (UInt window) const; // smooth a binarized fi of window
PixelF pixel (CoordD x, CoordD y) const; // bi-linear interpolation
PixelF pixel (CoordI x, CoordI y, UInt accuracy) const; // bi-linear interpolation using 2/4/8/16 quantization
PixelF pixel (const CSiteD& std) const {return pixel (std.x, std.y);} // bi-linear interpolation
PixelF pixel (const CSite& std, UInt accuracy) const {return pixel (std.x, std.y, accuracy);} // bi-linear interpolation
Double mean () const; // mean
Double mean (const CFloatImage* pfiMsk) const; // mean
Double sumAbs (const CRct& rct = CRct ()) const; // sum of absolute value
Double sumDeviation () const; // sum of first-order deviation
Double sumDeviation (const CFloatImage* pfiMsk) const; // sum of first-order deviation
Bool allValue (Float vl, const CRct& rct = CRct ()) const; // whether all pixels have value of vl
Bool biLevel (const CRct& rct = CRct ()) const; // whether it is a binary image
Bool atLeastOneValue (Float vl, const CRct& rct = CRct ()) const; // whether at least one pixel has value of vl
CRct whereVisible () const; // the tightest bounding box of non-transparent pixels
UInt numPixelsNotValued (Float vl, const CRct& rct = CRct ()) const; // number of pixels not valued vl in region rct
Double mse (const CFloatImage& fiCompare) const;
Double mse (const CFloatImage& fiCompare, const CFloatImage& fiMsk) const;
Double snr (const CFloatImage& fiCompare) const;
Double snr (const CFloatImage& fiCompare, const CFloatImage& fiMsk) const;
Void vdlDump (const Char* fileName = "tmp.vdl") const;
Void dump (FILE *) const;
Void txtDump (const Char* fileName = "tmp.txt") const;
Void txtDump (FILE *) const;
Void txtDumpMask (FILE *) const;
own CFloatImage* extendPadLeft (); // for block padding
own CFloatImage* extendPadRight (); // for block padding
own CFloatImage* extendPadTop (); // for block padding
own CFloatImage* extendPadDown (); // for block padding
// Overloaded operators
own CFloatImage& operator += (const CFloatImage &);
own CFloatImage* operator + (const CFloatImage& fi) const;
own CFloatImage* operator - (const CFloatImage& fi) const;
own CFloatImage* operator * (Float scale) const; // multiply every pixel by scale
own CFloatImage* operator / (Float scale) const; // divide every pixel by scale
own CFloatImage* operator * (const CTransform& tf) const;
own CFloatImage* operator * (const CAffine2D& aff) const {return warp (aff);} // affine warp
own CFloatImage* operator * (const CPerspective2D& persp) const {return warp (persp);} // perspective warp
Bool operator == (const CFloatImage& fi) const;
// Operations
Void where (const CRct& r); // crop based on r
Void pixel (CoordI x, CoordI y, PixelF p) {m_ppxlf [m_rc.offset (x, y)] = p;} // set a pixel
Void pixel (const CSite& st, PixelF p) {pixel (st.x, st.y, p);} // set a pixel
Void setRect (const CRct& rct); // set the rect
Void shift (Int nX, Int nY) {m_rc.shift (nX, nY);} // shift CRct by nX and nY
Void zeroPad (const CFloatImage* pfimsk); // zero padding
Void LPEPad (const CFloatImage* pfimsk); // Low Pass Extrapolation Padding
Void simplifiedRepeatPad(const CFloatImage *); // simplified repetitive padding
Void modSimplifiedRepeatPad(const CFloatImage *);
Void originalRepeatPad(const CFloatImage *);
Void modOriginalRepeatPad(const CFloatImage *);
Void repeatPad (const CFloatImage* pfimsk); // repetitive padding
Void sHorRepeatPad (const CFloatImage* pfimsk); // simplified horizontal repetitive padding
Void sVerRepeatPad (const CFloatImage* pfimsk); // simplified vertical repetitive padding
Void mrepeatPad (const CFloatImage* pfimsk); // modified repetitive padding
Void blockPad (const CFloatImage* pfiOrigMask, const UInt uintBlockSize); // block padding
Void dilatePad (const CFloatImage* pfiOrigMask); // morphological dilating padding
Void averagePad (const CFloatImage* pfiOrigMask); // averaging padding
Void repeatPadNew (const CFloatImage* pfimsk); // repetitive padding
Void threshold (Float thresh); // cut < thresh to be zero
Void binarize (Float fltThresh); // binarize it
Void checkRange (Float fltMin, Float fltMax);
Void overlay (const CFloatImage& fi); // overlay fi on top of *this
Void xorFi (const CFloatImage& fi); // xor two binarized fi
Void orFi (const CFloatImage& fi); // or of two binarized fi
Void andFi (const CFloatImage& fi); // or of two binarized fi
Void maskOut (const CFloatImage& fi); // mask out fi
Void mutiplyAlpha (const CFloatImage& fi); //multiply by alpha channel
Void cropOnAlpha (); // crop the image based on non-transparent value
Void quantize (int stepsize, Bool dpcm = FALSE);
// quantize the signal using the stepsize, dpcm if required
Void deQuantize (int stepsize, Bool dpcm = FALSE);
// deQuantize
Void operator = (const CFloatImage& fi);
Void roundNearestInt();
///////////////// implementation /////////////////
protected:
PixelF* m_ppxlf;
CRct m_rc;
Void allocate (const CRct& r, PixelF pxlf);
Void copyConstruct (const CFloatImage& fi, const CRct& rct);
Void swap (CFloatImage& fi);
Void fillColumnDownward (
const PixelF*& pxlOrigMask,
const PixelF*& ppxlOrigData,
PixelF*& ppxlFillMask,
PixelF*& ppxlFillData
) const;
Void fillColumnUpward (
const PixelF*& ppxlOrigMask,
const PixelF*& ppxlOrigData,
PixelF*& ppxlFillMask,
PixelF*& ppxlMaskData
) const;
Void fillRowRightward (
const PixelF*& ppxlOrigMask,
const PixelF*& ppxlOrigData,
PixelF*& ppxlFillMask,
PixelF*& ppxlFillData
) const;
Void fillRowLeftward (
const PixelF*& ppxlOrigMask,
const PixelF*& ppxlOrigData,
PixelF*& ppxlFillMask,
PixelF*& ppxlFillData
) const;
Void fillCorners (CRct& nBlkRct);
Void fillHoles (const CFloatImage *pfiFillMask);
own CFloatImage* smooth_ (UInt window) const; // smooth a binarized fi of window
// for block padding
Void dilation4 (CFloatImage* pMask);
Void dilation8 (CFloatImage* pMask);
};
#endif //_GRAYF_HPP_
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -