簡體   English   中英

更快地繪制帶有比例和 alpha 通道的 tbitmap

[英]Draw tbitmap with scale and alpha channel faster

下面的代碼復制一個大的位圖,將它與正確的背景混合,然后繪制一個帶有剪切區域的半透明圖像以節省繪制時間......圖像在一個數組中並預先縮放......

基於我對 C++ 和 Builder 圖形的有限知識,這已經通過了幾個級別的優化......

編輯:更新代碼... blend();

void blend(Graphics::TBitmap *dst,int x,int y,Graphics::TBitmap *src,BYTE 
alpha)
{
const int n=3;          // pixel align [Bytes]
int dx0,dy0,dx1,dy1,    // dst BBOX
    sx0,sy0,sx1,sy1,    // src BBOX
    dx,dy,sx,sy,i;
BYTE *dp,*sp;
WORD a,_a,sc,dc,da[256],sa[256];

// compute BBOX (handle clipping)
dx=src->Width; dy=src->Height;
dx0=x; sx0=0; dx1=x+dx; sx1=dx;
dy0=y; sy0=0; dy1=y+dy; sy1=dy;


// blend
a=alpha; _a=255-a;
for (i=0;i<256;i++){ da[i]=_a*i; sa[i]=a*i; }   // precompute BYTE*a and 
BYTE*_a LUTs

for (dy=dy0,sy=sy0;dy<dy1;dy++,sy++)        // ScanLines
    {
    dp=(BYTE*)dst->ScanLine[dy]+(n*dx0);
    sp=(BYTE*)src->ScanLine[sy]+(n*sx0);
    for (dx=dx0,sx=sx0;dx<dx1;dx++,sx++)    // single ScanLine
     for (i=0;i<n;i++,dp++,sp++)            // RGB
      *dp=WORD((sa[*sp]+da[*dp])>>8);       // blend function
    }
}

//--------------------------------------------------------------------------

    det1maps.push_back( new Graphics::TBitmap() );
    for (int i = 1; i < 176; i++)
    {
        det1maps.push_back( new Graphics::TBitmap() );
        det1maps[i]->SetSize(t,t);
        det1maps[i]->Canvas->StretchDraw(Rect(0, 0, t, t), Det1_bmp.get()); // scale
        t = t + 24;
    }

//----------------- 編輯 3 當前版本 1/18

det1maps[ss]->Transparent = true;
Form1->imgTemp->Picture->Assign(layer0_bmap.get()); //why background first?
HRGN MyRgn;
MyRgn = ::CreateRectRgn(0,0,Sw,Sh);
::SelectClipRgn(Form1->imgTemp->Canvas->Handle,MyRgn); //clip

Form1->imgTemp->Canvas->Draw(X3,Y3,det1maps[ss]); // draw det

blend(layer0_bmap.get(),0,0,Form1->imgTemp->Picture->Bitmap,int(obj[index]));

這是我剛剛放在一起的小型簡單C++/VCL ScanLine Alpha Blend示例:

//---------------------------------------------------------------------------
void blend(Graphics::TBitmap *dst,int x,int y,Graphics::TBitmap *src,BYTE alpha)
    {
    const int n=3;          // pixel align [Bytes]
    int dx0,dy0,dx1,dy1,    // dst BBOX
        sx0,sy0,sx1,sy1,    // src BBOX
        dx,dy,sx,sy,i;
    BYTE *dp,*sp;
    WORD a,_a,sc,dc,da[256],sa[256];
    // compute BBOX (handle clipping)
    dx=src->Width; dy=src->Height;
    dx0=x; sx0=0; dx1=x+dx; sx1=dx;
    dy0=y; sy0=0; dy1=y+dy; sy1=dy;
    if (dx0<0){ sx0-=dx0; dx0=0; }
    if (dy0<0){ sy0-=dy0; dy0=0; }
    dx=dst->Width; dy=dst->Height;
    if (dx1>dx){ sx1+=dx-dx1; dx1=dx; }
    if (dy1>dy){ sy1+=dy-dy1; dy1=dy; }
    // make sure config is compatible with ScanLine[]
    dst->HandleType=bmDIB; dst->PixelFormat=pf24bit;
    src->HandleType=bmDIB; src->PixelFormat=pf24bit;
    // blend
    a=alpha; _a=255-a;
    for (i=0;i<256;i++){ da[i]=_a*i; sa[i]=a*i; }   // precompite BYTE*a and BYTE*_a LUTs
    for (dy=dy0,sy=sy0;dy<dy1;dy++,sy++)        // ScanLines
        {
        dp=(BYTE*)dst->ScanLine[dy]+(n*dx0);
        sp=(BYTE*)src->ScanLine[sy]+(n*sx0);
        for (dx=dx0,sx=sx0;dx<dx1;dx++,sx++)    // single ScanLine
         for (i=0;i<n;i++,dp++,sp++)            // RGB
          *dp=WORD((sa[*sp]+da[*dp])>>8);       // blend function
        }
    }
//---------------------------------------------------------------------------

我只是在每個像素/通道的基礎上處理圖像,並為每個通道 (R,G,B) 計算:

dst_pixel =  ( src_pixel*alpha + dst_pixel*(255-alpha) )/255

其中通道和 alpha 是 8 位無符號整數......為了速度,我使用了 24 位像素格式(通常我使用 32 位代替)。

為了避免*,/在混合中,我預先計算了 2 個LUT ,其中包含number*alphanumber*(255-alpha)所有可能組合。 除法是通過位移>>8

為了提高速度,您可以將dst圖像的所有ScanLine[]記住一次,然后將其用作目標圖像將被多次使用...

當我將 2 個1024x768圖像混合在一起進行測試時,我的設置花費了<=9ms 最慢的操作是ScanLine[]訪問,並且在混合之前將圖像格式化為像素格式...

這是 GIF 預覽(縮小 1/4 並由我的捕獲器抖動,因此它適合 imgur 2MByte 限制):

動畫片

這是我用於此的代碼(單定時器 VCL 應用程序):

//$$---- Form CPP ----
//---------------------------------------------------------------------------
#include <vcl.h>
#pragma hdrstop
#include "win_main.h"
#include <math.h>
#include <jpeg.hpp>
//---------------------------------------------------------------------------
#pragma package(smart_init)
#pragma resource "*.dfm"
TMain *Main;
Graphics::TBitmap *bmp,*bmp0,*bmp1; // back buffer, image0, image1, ...
//---------------------------------------------------------------------------
void blend(Graphics::TBitmap *dst,int x,int y,Graphics::TBitmap *src,BYTE alpha)
    {
    const int n=3;          // pixel align [Bytes]
    int dx0,dy0,dx1,dy1,    // dst BBOX
        sx0,sy0,sx1,sy1,    // src BBOX
        dx,dy,sx,sy,i;
    BYTE *dp,*sp;
    WORD a,_a,sc,dc,da[256],sa[256];
    // compute BBOX (handle clipping)
    dx=src->Width; dy=src->Height;
    dx0=x; sx0=0; dx1=x+dx; sx1=dx;
    dy0=y; sy0=0; dy1=y+dy; sy1=dy;
    if (dx0<0){ sx0-=dx0; dx0=0; }
    if (dy0<0){ sy0-=dy0; dy0=0; }
    dx=dst->Width; dy=dst->Height;
    if (dx1>dx){ sx1+=dx-dx1; dx1=dx; }
    if (dy1>dy){ sy1+=dy-dy1; dy1=dy; }
    // make sure config is compatible with ScanLine[]
    dst->HandleType=bmDIB; dst->PixelFormat=pf24bit;
    src->HandleType=bmDIB; src->PixelFormat=pf24bit;
    // blend
    a=alpha; _a=255-a;
    for (i=0;i<256;i++){ da[i]=_a*i; sa[i]=a*i; }   // precompite BYTE*a and BYTE*_a LUTs
    for (dy=dy0,sy=sy0;dy<dy1;dy++,sy++)        // ScanLines
        {
        dp=(BYTE*)dst->ScanLine[dy]+(n*dx0);
        sp=(BYTE*)src->ScanLine[sy]+(n*sx0);
        for (dx=dx0,sx=sx0;dx<dx1;dx++,sx++)    // single ScanLine
         for (i=0;i<n;i++,dp++,sp++)            // RGB
          *dp=WORD((sa[*sp]+da[*dp])>>8);       // blend function
        }
    }
//---------------------------------------------------------------------------
//---------------------------------------------------------------------------
//---------------------------------------------------------------------------
void TMain::draw()
    {
    bmp->Canvas->Draw(0,0,bmp0);            // render background bmp0
    static float a=0.0; a+=0.025*M_PI;
    blend(bmp,0,0,bmp1,fabs(255.0*sin(a))); // alfa blend in bmp1
    Main->Canvas->Draw(0,0,bmp);            // show result on screen
    }
//---------------------------------------------------------------------------
__fastcall TMain::TMain(TComponent* Owner) : TForm(Owner)
    {
    // create bitmaps
    bmp=new Graphics::TBitmap;
    bmp0=new Graphics::TBitmap;
    bmp1=new Graphics::TBitmap;
    // laod images
    TJPEGImage *jpg=new TJPEGImage;
    jpg->LoadFromFile("img0.jpg"); bmp0->Assign(jpg);
    jpg->LoadFromFile("img1.jpg"); bmp1->Assign(jpg);
    delete jpg;
    }
//---------------------------------------------------------------------------
void __fastcall TMain::FormDestroy(TObject *Sender)
    {
    // delete bitmaps
    delete bmp0;
    delete bmp1;
    delete bmp;
    }
//---------------------------------------------------------------------------
void __fastcall TMain::FormResize(TObject *Sender)
    {
    bmp->Width =ClientWidth;
    bmp->Height=ClientHeight;
    }
//---------------------------------------------------------------------------
void __fastcall TMain::FormPaint(TObject *Sender)
    {
    draw();
    }
//---------------------------------------------------------------------------
void __fastcall TMain::tim_redrawTimer(TObject *Sender)
    {
    draw();
    }
//---------------------------------------------------------------------------

這里是圖片(我在谷歌圖片上找到的第一個漂亮的 1024x768 圖片):

img0

圖片1

這里預覽混合結果:

混合

有關ScanLine 的更多信息,請參閱:

如果您需要更高的速度,那么您應該選擇GPU Blending( OpenGLDirectX )。

[Edit2] 數組 + 矩形示例

在您編輯您的問題后,現在很明顯:

  1. 您的位圖數組根本不是數組

    它是某種列表模板,如vector<Graphics::TBitmap*>或類似的......所以你不能像我一樣訪問 bmps 的線性數組。 為了讓您的生活更輕松,我使用了具有類似屬性的我的模板,以便您可以了解如何處理這些模板(抱歉,我無法共享模板代碼,但您只需要將List<T>更改為Vector<T>或您正在使用的任何內容。 ..

    這就是數組指針對您不起作用的原因,因為您沒有。 它可能是您的模板向某個成員公開它。 我的像map.dat一樣,所以如果不是線性存儲,你的可能有類似的東西或根本沒有。

  2. 您只混合了 2 個圖像而不是整個陣列

    因此您可以使用第一個示例並添加 ScanLine 預加載,因為您的圖像是靜態的...對后台緩沖區圖像執行相同操作,因為只有在調整大小后才會更改。

當我把所有東西放在一起時,結果是:

//$$---- Form CPP ----
//---------------------------------------------------------------------------
#include <vcl.h>
#pragma hdrstop
#include "win_main.h"
#include <math.h>
#include <jpeg.hpp>
#include "list.h"           // mine list<T> template you got probably vector<> or something similar instead
#include "performance.h"    // this is mine tbeg/tend/tstr time measurement
//---------------------------------------------------------------------------
#pragma package(smart_init)
#pragma resource "*.dfm"
TMain *Main;
//---------------------------------------------------------------------------
// [back buffer]
Graphics::TBitmap *bmp;             // bitmap
BYTE **bmp_pyx=NULL;                // preloaded ScanLines [y][x]
void bmp_init()                     // create preloaded ScanLines
    {
    bmp_pyx=new BYTE*[bmp->Height];
    for (int y=0;y<bmp->Height;y++)
     bmp_pyx[y]=(BYTE*)bmp->ScanLine[y];
    }
void bmp_exit()                     // release preloaded ScanLines
    {
    delete[] bmp_pyx;
    }
//---------------------------------------------------------------------------
// [array of images]
const AnsiString filename[]=        // filenames
    {
    "img0.jpg",
    "img1.jpg",
    "img2.jpg",
    "img3.jpg",
    "img4.jpg",
    "img5.jpg",
    "img6.jpg",
    "img7.jpg",
    "img8.jpg",
    "img9.jpg",
    ""
    };
List<Graphics::TBitmap*> map;       // your "array" of bitmaps
int maps=0;                         // number of images
BYTE ***map_pyx=NULL;               // preloaded ScanLines [ix][y][x]
//---------------------------------------------------------------------------
void map_init()                     // alocate and prepare data
    {
    int i,y;
    Graphics::TBitmap *bmp;
    TJPEGImage *jpg=new TJPEGImage;
    // create "array" of bmp (you already got this)
    for (maps=0;filename[maps]!="";maps++)
        {
        map.add(new Graphics::TBitmap); // this is like your push_back(new Graphics::TBitmap)
        jpg->LoadFromFile(filename[maps]);  // filename[] -> jpg -> bmp -> map[]
        map[maps]->Assign(jpg);             // here you can also rescale or whatever you want to do...
        map[maps]->HandleType=bmDIB;
        map[maps]->PixelFormat=pf24bit;
        }
    // create preloaded ScanLines (you need to add this into your app init)
    map_pyx=new BYTE**[maps];                   // **map_pyx[]
    for (i=0;i<maps;i++)
        {
        map_pyx[i]=new BYTE*[map[i]->Height];   // *map_pyx[][]
        for (y=0;y<map[i]->Height;y++)          // map_pyx[][]]
         map_pyx[i][y]=(BYTE*)map[i]->ScanLine[y];
        }
    delete jpg;
    }
//---------------------------------------------------------------------------
void map_exit()                     // release data (you need to add this in app exit)
    {
    int i;
    for (i=0;i<maps;i++)
        {
        delete   map[i];
        delete[] map_pyx[i];
        }
    delete[] map_pyx;
    }
//---------------------------------------------------------------------------
void blend_rec(BYTE **dp,int x0,int y0,int x1,int y1,BYTE **sp,BYTE alpha)
    {
    const int n=3;          // pixel align [Bytes]
    int x,y,i;
    BYTE *d,*s;
    WORD da[256],sa[256];
    // pixelformat align
    x0*=n; x1*=n;
    // prepare alpha*BYTE and (255-alpha)*BYTE LUTs
    y=    alpha; for (x=0;x<256;x++) sa[x]=x*y;
    y=255-alpha; for (x=0;x<256;x++) da[x]=x*y;
    // blend
    for (y=y0;y<y1;y++)
        {
        d=dp[y]+x0;
        s=sp[y]+x0;
        for (x=x0;x<x1;x++,d++,s++)
         *d=WORD((sa[*s]+da[*d])>>8);       // blend function
        }
    // release data
    }
//---------------------------------------------------------------------------
void TMain::draw()
    {
    bmp->Canvas->Draw(0,0,map[0]);              // render background bmp[0]
    static float a=0.0; a+=0.025*M_PI;          // animation ...
    BYTE alpha=128+float(127.0*sin(a));
    tbeg();
    blend_rec(bmp_pyx,200,500,400,600,map_pyx[1],alpha);    // add the blended rectangle (except background which is bmp[0]
    tend(); Caption=tstr();
    Canvas->Draw(0,0,bmp);                      // show on screen
//  bmp->SaveToFile("out.bmp");
    }
//---------------------------------------------------------------------------
__fastcall TMain::TMain(TComponent* Owner) : TForm(Owner)
    {
    // create bitmaps
    bmp=new Graphics::TBitmap;
    bmp_init();
    map_init();
    }
//---------------------------------------------------------------------------
void __fastcall TMain::FormDestroy(TObject *Sender)
    {
    // delete bitmaps
    delete bmp;
    bmp_exit();
    map_exit();
    }
//---------------------------------------------------------------------------
void __fastcall TMain::FormResize(TObject *Sender)
    {
    bmp->Width =ClientWidth;
    bmp->Height=ClientHeight;
    bmp->HandleType=bmDIB;
    bmp->PixelFormat=pf24bit;
    bmp_exit();
    bmp_init();
    }
//---------------------------------------------------------------------------
void __fastcall TMain::FormPaint(TObject *Sender)
    {
    draw();
    }
//---------------------------------------------------------------------------
void __fastcall TMain::tim_redrawTimer(TObject *Sender)
    {
    draw();
    }
//---------------------------------------------------------------------------

在我選擇的矩形設置中,混合在不到0.5ms時間內完成。 正如你所看到的,它比原來的9ms更快......因為如果你使用剪輯區域,你仍然會混合整個圖像而不是復制結果。 這種方法只混合和復制需要的東西。

當心我刪除了范圍檢查,因此請確保矩形在圖像內...

如果您想以與我相同的方式測量時間,我將使用我的以下代碼:

性能.h:

//---------------------------------------------------------------------------
//--- Performance counter time measurement: 2.01 ----------------------------
//---------------------------------------------------------------------------
#ifndef _performance_h
#define _performance_h
//---------------------------------------------------------------------------
const int   performance_max=64;                 // push urovni
double      performance_Tms=-1.0,               // perioda citaca [ms]
            performance_tms=0.0,                // zmerany cas po tend [ms]
            performance_t0[performance_max];    // zmerane start casy [ms]
int         performance_ix=-1;                  // index aktualneho casu
//---------------------------------------------------------------------------
void tbeg(double *t0=NULL)  // mesure start time
    {
    double t;
    LARGE_INTEGER i;
    if (performance_Tms<=0.0)
        {
        for (int j=0;j<performance_max;j++) performance_t0[j]=0.0;
        QueryPerformanceFrequency(&i); performance_Tms=1000.0/double(i.QuadPart);
        }
    QueryPerformanceCounter(&i); t=double(i.QuadPart); t*=performance_Tms;
    if (t0) { t0[0]=t; return; }
    performance_ix++;
    if ((performance_ix>=0)&&(performance_ix<performance_max)) performance_t0[performance_ix]=t;
    }
//---------------------------------------------------------------------------
void tpause(double *t0=NULL)    // stop counting time between tbeg()..tend() calls
    {
    double t;
    LARGE_INTEGER i;
    QueryPerformanceCounter(&i); t=double(i.QuadPart); t*=performance_Tms;
    if (t0) { t0[0]=t-t0[0]; return; }
    if ((performance_ix>=0)&&(performance_ix<performance_max)) performance_t0[performance_ix]=t-performance_t0[performance_ix];
    }
//---------------------------------------------------------------------------
void tresume(double *t0=NULL)   // resume counting time between tbeg()..tend() calls
    {
    double t;
    LARGE_INTEGER i;
    QueryPerformanceCounter(&i); t=double(i.QuadPart); t*=performance_Tms;
    if (t0) { t0[0]=t-t0[0]; return; }
    if ((performance_ix>=0)&&(performance_ix<performance_max)) performance_t0[performance_ix]=t-performance_t0[performance_ix];
    }
//---------------------------------------------------------------------------
double tend(double *t0=NULL)    // return duration [ms] between matching tbeg()..tend() calls
    {
    double t;
    LARGE_INTEGER i;
    QueryPerformanceCounter(&i); t=double(i.QuadPart); t*=performance_Tms;
    if (t0) { t-=t0[0]; performance_tms=t; return t; }
    if ((performance_ix>=0)&&(performance_ix<performance_max)) t-=performance_t0[performance_ix]; else t=0.0;
    performance_ix--;
    performance_tms=t;
    return t;
    }
//---------------------------------------------------------------------------
double tper(double *t0=NULL)    // return duration [ms] between tper() calls
    {
    double t,tt;
    LARGE_INTEGER i;
    if (performance_Tms<=0.0)
        {
        for (int j=0;j<performance_max;j++) performance_t0[j]=0.0;
        QueryPerformanceFrequency(&i); performance_Tms=1000.0/double(i.QuadPart);
        }
    QueryPerformanceCounter(&i); t=double(i.QuadPart); t*=performance_Tms;
    if (t0) { tt=t-t0[0]; t0[0]=t; performance_tms=tt; return tt; }
    performance_ix++;
    if ((performance_ix>=0)&&(performance_ix<performance_max))
        {
        tt=t-performance_t0[performance_ix];
        performance_t0[performance_ix]=t;
        }
    else { t=0.0; tt=0.0; };
    performance_ix--;
    performance_tms=tt;
    return tt;
    }
//---------------------------------------------------------------------------
AnsiString tstr()
    {
    AnsiString s;
    s=s.sprintf("%8.3lf",performance_tms); while (s.Length()<8) s=" "+s; s="["+s+" ms]";
    return s;
    }
//---------------------------------------------------------------------------
AnsiString tstr(int N)
    {
    AnsiString s;
    s=s.sprintf("%8.3lf",performance_tms/double(N)); while (s.Length()<8) s=" "+s; s="["+s+" ms]";
    return s;
    }
//---------------------------------------------------------------------------
//---------------------------------------------------------------------------
#endif
//---------------------------------------------------------------------------
//---------------------------------------------------------------------------

暫無
暫無

聲明:本站的技術帖子網頁,遵循CC BY-SA 4.0協議,如果您需要轉載,請注明本站網址或者原文地址。任何問題請咨詢:yoyou2525@163.com.

 
粵ICP備18138465號  © 2020-2024 STACKOOM.COM