old-nall/image/utility.hpp
2016-05-15 12:42:10 +02:00

118 lines
4.4 KiB
C++

#pragma once
namespace nall {
auto image::crop(unsigned outputX, unsigned outputY, unsigned outputWidth, unsigned outputHeight) -> bool {
if(outputX + outputWidth > _width) return false;
if(outputY + outputHeight > _height) return false;
uint8_t* outputData = allocate(outputWidth, outputHeight, stride());
unsigned outputPitch = outputWidth * stride();
#pragma omp parallel for
for(unsigned y = 0; y < outputHeight; y++) {
const uint8_t* sp = _data + pitch() * (outputY + y) + stride() * outputX;
uint8_t* dp = outputData + outputPitch * y;
for(unsigned x = 0; x < outputWidth; x++) {
write(dp, read(sp));
sp += stride();
dp += stride();
}
}
delete[] _data;
_data = outputData;
_width = outputWidth;
_height = outputHeight;
return true;
}
auto image::alphaBlend(uint64_t alphaColor) -> void {
uint64_t alphaR = (alphaColor & _red.mask() ) >> _red.shift();
uint64_t alphaG = (alphaColor & _green.mask()) >> _green.shift();
uint64_t alphaB = (alphaColor & _blue.mask() ) >> _blue.shift();
#pragma omp parallel for
for(unsigned y = 0; y < _height; y++) {
uint8_t* dp = _data + pitch() * y;
for(unsigned x = 0; x < _width; x++) {
uint64_t color = read(dp);
uint64_t colorA = (color & _alpha.mask()) >> _alpha.shift();
uint64_t colorR = (color & _red.mask() ) >> _red.shift();
uint64_t colorG = (color & _green.mask()) >> _green.shift();
uint64_t colorB = (color & _blue.mask() ) >> _blue.shift();
double alphaScale = (double)colorA / (double)((1 << _alpha.depth()) - 1);
colorA = (1 << _alpha.depth()) - 1;
colorR = (colorR * alphaScale) + (alphaR * (1.0 - alphaScale));
colorG = (colorG * alphaScale) + (alphaG * (1.0 - alphaScale));
colorB = (colorB * alphaScale) + (alphaB * (1.0 - alphaScale));
write(dp, (colorA << _alpha.shift()) | (colorR << _red.shift()) | (colorG << _green.shift()) | (colorB << _blue.shift()));
dp += stride();
}
}
}
auto image::alphaMultiply() -> void {
unsigned divisor = (1 << _alpha.depth()) - 1;
#pragma omp parallel for
for(unsigned y = 0; y < _height; y++) {
uint8_t* dp = _data + pitch() * y;
for(unsigned x = 0; x < _width; x++) {
uint64_t color = read(dp);
uint64_t colorA = (color & _alpha.mask()) >> _alpha.shift();
uint64_t colorR = (color & _red.mask() ) >> _red.shift();
uint64_t colorG = (color & _green.mask()) >> _green.shift();
uint64_t colorB = (color & _blue.mask() ) >> _blue.shift();
colorR = (colorR * colorA) / divisor;
colorG = (colorG * colorA) / divisor;
colorB = (colorB * colorA) / divisor;
write(dp, (colorA << _alpha.shift()) | (colorR << _red.shift()) | (colorG << _green.shift()) | (colorB << _blue.shift()));
dp += stride();
}
}
}
auto image::transform(const image& source) -> void {
return transform(source._endian, source._depth, source._alpha.mask(), source._red.mask(), source._green.mask(), source._blue.mask());
}
auto image::transform(bool outputEndian, unsigned outputDepth, uint64_t outputAlphaMask, uint64_t outputRedMask, uint64_t outputGreenMask, uint64_t outputBlueMask) -> void {
if(_endian == outputEndian && _depth == outputDepth && _alpha.mask() == outputAlphaMask && _red.mask() == outputRedMask && _green.mask() == outputGreenMask && _blue.mask() == outputBlueMask) return;
image output(outputEndian, outputDepth, outputAlphaMask, outputRedMask, outputGreenMask, outputBlueMask);
output.allocate(_width, _height);
#pragma omp parallel for
for(unsigned y = 0; y < _height; y++) {
const uint8_t* sp = _data + pitch() * y;
uint8_t* dp = output._data + output.pitch() * y;
for(unsigned x = 0; x < _width; x++) {
uint64_t color = read(sp);
sp += stride();
uint64_t a = (color & _alpha.mask()) >> _alpha.shift();
uint64_t r = (color & _red.mask() ) >> _red.shift();
uint64_t g = (color & _green.mask()) >> _green.shift();
uint64_t b = (color & _blue.mask() ) >> _blue.shift();
a = normalize(a, _alpha.depth(), output._alpha.depth());
r = normalize(r, _red.depth(), output._red.depth());
g = normalize(g, _green.depth(), output._green.depth());
b = normalize(b, _blue.depth(), output._blue.depth());
output.write(dp, (a << output._alpha.shift()) | (r << output._red.shift()) | (g << output._green.shift()) | (b << output._blue.shift()));
dp += output.stride();
}
}
operator=(move(output));
}
}