更快地绘制具有比例和 alpha 通道的 tbitmap
Draw tbitmap with scale and alpha channel faster
以下代码复制一个大位图，将其与正确的背景混合，然后绘制一个带有裁剪区域的半透明图像以节省绘制时间...图像在数组中并预先缩放...
基于我对 C++ 和 Builder 图形的有限了解，这已通过多个级别的优化...
编辑： 更新代码... 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;
}
//----------------编辑 3 当前版本 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]));
这里是简单的小C++/VCLScanLineAlpha Blend例子我刚拼起来的:
//---------------------------------------------------------------------------
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 pixel/channel 基础上处理图像，并为每个通道 (R,G,B) 计算：
dst_pixel = ( src_pixel*alpha + dst_pixel*(255-alpha) )/255
其中通道和 alpha 是 8 位无符号整数...为了提高速度，我使用 24 位像素格式（通常我使用 32 位）。
为了避免在混合中使用 *,/，我预先计算了 2 个 LUT，其中包含 number*alpha 和 number*(255-alpha) 的所有可能组合。除法是通过移位 >>8.
为了提高速度，您可以将 dst 图像的所有 ScanLine[] 记住一次到您的数组中，然后使用它作为目标图像将被多次使用 ...
当我将 2 1024x768 图像混合在一起进行测试时，我的设置花费了 <=9ms。最慢的操作是 ScanLine[] 访问，图像在混合之前被格式化为像素格式...
此处为 GIF 预览图（缩小 1/4 并由我的捕捉器抖动以适合 imgur 2MByte 限制）：
这是我为此使用的代码（单计时器 VCL 应用程序）：
//$$---- 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();
}
//---------------------------------------------------------------------------
这里是图片（我在 Google 图片上找到的第一张漂亮的 1024x768 图片）：
这里是混合结果预览：
有关 ScanLine 的更多信息，请参阅：
- gfx rendering
如果您需要更快的速度，那么您应该选择 GPU 混合（OpenGL 或 DirectX).
[Edit2]数组+矩形例子
在你编辑你的问题之后，它现在很明显了：
你的位图数组根本不是数组
它更像是某种列表模板，例如 vector<Graphics::TBitmap*> 或类似的...所以您无法像我一样访问 bmp 的线性数组。为了让您的生活更轻松，我使用了具有相似属性的我的模板，这样您就可以看到如何处理这些（抱歉，我不能共享模板代码，但您只需要将 List<T> 更改为 Vector<T> 或任何您想要的使用 ...
这就是为什么数组指针对你不起作用的原因，因为你没有。它可能是您的模板与某些成员公开的。我的像 map.dat 这样，如果不是线性存储，你的可能有类似的东西或根本没有。
您只混合了 2 个图像而不是整个阵列
因此您可以使用第一个示例并添加 ScanLine 预加载，因为您的图像是静态的...对后台缓冲区图像执行相同操作，因为它仅在调整大小后才会更改。
当我把所有的东西放在一起时，结果是：
//$$---- 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();
}
//---------------------------------------------------------------------------
在我选择的矩形设置中，混合在不到 0.5ms 的时间内完成。如您所见，它比原始 9ms 更快...因为如果您使用的是剪辑区域，您仍然会混合整个图像，只是不会复制结果。此方法仅混合和复制需要的内容。
注意我删除了范围检查，因此请确保矩形在图像内部...
如果你想像我一样测量时间，我正在使用我的这个代码：
Performance.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
//---------------------------------------------------------------------------
//---------------------------------------------------------------------------
以下代码复制一个大位图，将其与正确的背景混合，然后绘制一个带有裁剪区域的半透明图像以节省绘制时间...图像在数组中并预先缩放...
基于我对 C++ 和 Builder 图形的有限了解，这已通过多个级别的优化...
编辑： 更新代码... 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;
}
//----------------编辑 3 当前版本 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]));
这里是简单的小C++/VCLScanLineAlpha Blend例子我刚拼起来的:
//---------------------------------------------------------------------------
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 pixel/channel 基础上处理图像，并为每个通道 (R,G,B) 计算：
dst_pixel = ( src_pixel*alpha + dst_pixel*(255-alpha) )/255
其中通道和 alpha 是 8 位无符号整数...为了提高速度，我使用 24 位像素格式（通常我使用 32 位）。
为了避免在混合中使用 *,/，我预先计算了 2 个 LUT，其中包含 number*alpha 和 number*(255-alpha) 的所有可能组合。除法是通过移位 >>8.
为了提高速度，您可以将 dst 图像的所有 ScanLine[] 记住一次到您的数组中，然后使用它作为目标图像将被多次使用 ...
当我将 2 1024x768 图像混合在一起进行测试时，我的设置花费了 <=9ms。最慢的操作是 ScanLine[] 访问，图像在混合之前被格式化为像素格式...
此处为 GIF 预览图（缩小 1/4 并由我的捕捉器抖动以适合 imgur 2MByte 限制）：
这是我为此使用的代码（单计时器 VCL 应用程序）：
//$$---- 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();
}
//---------------------------------------------------------------------------
这里是图片（我在 Google 图片上找到的第一张漂亮的 1024x768 图片）：
这里是混合结果预览：
有关 ScanLine 的更多信息，请参阅：
- gfx rendering
如果您需要更快的速度，那么您应该选择 GPU 混合（OpenGL 或 DirectX).
[Edit2]数组+矩形例子
在你编辑你的问题之后，它现在很明显了：
你的位图数组根本不是数组
它更像是某种列表模板，例如
vector<Graphics::TBitmap*>或类似的...所以您无法像我一样访问 bmp 的线性数组。为了让您的生活更轻松，我使用了具有相似属性的我的模板，这样您就可以看到如何处理这些（抱歉，我不能共享模板代码，但您只需要将List<T>更改为Vector<T>或任何您想要的使用 ...这就是为什么数组指针对你不起作用的原因，因为你没有。它可能是您的模板与某些成员公开的。我的像
map.dat这样，如果不是线性存储，你的可能有类似的东西或根本没有。您只混合了 2 个图像而不是整个阵列
因此您可以使用第一个示例并添加 ScanLine 预加载，因为您的图像是静态的...对后台缓冲区图像执行相同操作，因为它仅在调整大小后才会更改。
当我把所有的东西放在一起时，结果是：
//$$---- 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();
}
//---------------------------------------------------------------------------
在我选择的矩形设置中，混合在不到 0.5ms 的时间内完成。如您所见，它比原始 9ms 更快...因为如果您使用的是剪辑区域，您仍然会混合整个图像，只是不会复制结果。此方法仅混合和复制需要的内容。
注意我删除了范围检查，因此请确保矩形在图像内部...
如果你想像我一样测量时间，我正在使用我的这个代码：
Performance.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
//---------------------------------------------------------------------------
//---------------------------------------------------------------------------