效果图
CLAHE简介
HE 直方图增强,大家都不陌生,是一种比较古老的对比度增强算法,它有两种变体:AHE 和 CLAHE;两者都是自适应的增强算法,功能差不多,但是前者有一个很大的缺陷,就是有时候会过度放大图像中相同区域的噪声,为了解决这一问题,出现了 HE 的另一种改进算法,就是 CLAHE;CLAHE 是另外一种直方图均衡算法,CLAHE 和 AHE 的区别在于前者对区域对比度实行了限制,并且利用插值来加快计算。它能有效的增强或改善图像(局部)对比度,从而获取更多图像相关边缘信息有利于分割。还能够有效改善 AHE 中放大噪声的问题。另外,CLAHE 的有一个用途是被用来对图像去雾。
详细理论请参考博客
OpenCV源码的本地路径: %OPENCV%\opencv\sources\modules\imgproc\src\clahe.cpp
clahe.cpp
// ----------------------------------------------------------------------
// CLAHE
namespace
{
class CLAHE_CalcLut_Body : public cv::ParallelLoopBody
{
public:
CLAHE_CalcLut_Body(const cv::Mat& src, cv::Mat& lut, cv::Size tileSize, int tilesX, int clipLimit, float lutScale) :
src_(src), lut_(lut), tileSize_(tileSize), tilesX_(tilesX), clipLimit_(clipLimit), lutScale_(lutScale)
{
}
void operator ()(const cv::Range& range) const;
private:
cv::Mat src_;
mutable cv::Mat lut_;
cv::Size tileSize_;
int tilesX_;
int clipLimit_;
float lutScale_;
};
// 计算直方图查找表
void CLAHE_CalcLut_Body::operator ()(const cv::Range& range) const
{
const int histSize = 256;
uchar* tileLut = lut_.ptr(range.start);
const size_t lut_step = lut_.step; // size = tilesX_*tilesY_ * lut_step
// Range(0, tilesX_ * tilesY_),全图被分为tilesX_*tiles_Y个块
for (int k = range.start; k < range.end; ++k, tileLut += lut_step)
{
// (tx, ty)表示当前所在是哪一块
// (0, 0) (1, 0)...(tilesX_-1, 0)
// (0, 1) (1, 1)...(tilesX_-1, 1)
// ...
// (0, tilesY_-1)... (tilesX_-1, tilesY_-1)
const int ty = k / tilesX_;
const int tx = k % tilesX_;
// retrieve tile submatrix
// 注意:tileSize.width表示分块的宽度,tileSize.height表示分块高度
cv::Rect tileROI;
tileROI.x = tx * tileSize_.width; // 换算为全局坐标
tileROI.y = ty * tileSize_.height;
tileROI.width = tileSize_.width;
tileROI.height = tileSize_.height;
const cv::Mat tile = src_(tileROI);
// calc histogram
int tileHist[histSize] = { 0, };
// 统计 ROI 的直方图
int height = tileROI.height;
const size_t sstep = tile.step;
for (const uchar* ptr = tile.ptr<uchar>(0); height--; ptr += sstep)
{
int x = 0;
for (; x <= tileROI.width - 4; x += 4)
{
int t0 = ptr[x], t1 = ptr[x + 1];
tileHist[t0]++; tileHist[t1]++;
t0 = ptr[x + 2]; t1 = ptr[x + 3];
tileHist[t0]++; tileHist[t1]++;
}
for (; x < tileROI.width; ++x)
tileHist[ptr[x]]++;
}
// clip histogram
if (clipLimit_ > 0)
{
// how many pixels were clipped
int clipped = 0;
for (int i = 0; i < histSize; ++i)
{
// 超过裁剪阈值
if (tileHist[i] > clipLimit_)
{
clipped += tileHist[i] - clipLimit_;
tileHist[i] = clipLimit_;
}
}
// redistribute clipped pixels
int redistBatch = clipped / histSize;
int residual = clipped - redistBatch * histSize;
// 平均分配裁剪的差值到所有直方图
for (int i = 0; i < histSize; ++i)
tileHist[i] += redistBatch;
// 处理差值的余数
for (int i = 0; i < residual; ++i)
tileHist[i]++;
}
// calc Lut
int sum = 0;
for (int i = 0; i < histSize; ++i)
{
// 累加直方图
sum += tileHist[i];
tileLut[i] = cv::saturate_cast<uchar>(sum * lutScale_); // static_cast<float>(histSize - 1) / tileSizeTotal
}
}
}
class CLAHE_Interpolation_Body : public cv::ParallelLoopBody
{
public:
CLAHE_Interpolation_Body(const cv::Mat& src, cv::Mat& dst, const cv::Mat& lut, cv::Size tileSize, int tilesX, int tilesY) :
src_(src), dst_(dst), lut_(lut), tileSize_(tileSize), tilesX_(tilesX), tilesY_(tilesY)
{
}
void operator ()(const cv::Range& range) const;
private:
cv::Mat src_;
mutable cv::Mat dst_;
cv::Mat lut_;
cv::Size tileSize_;
int tilesX_;
int tilesY_;
};
// 根据相邻4块的直方图插值
void CLAHE_Interpolation_Body::operator ()(const cv::Range& range) const
{
const size_t lut_step = lut_.step;
// Range(0, src.rows)
for (int y = range.start; y < range.end; ++y)
{
const uchar* srcRow = src_.ptr<uchar>(y);
uchar* dstRow = dst_.ptr<uchar>(y);
const float tyf = (static_cast<float>(y) / tileSize_.height) - 0.5f;
int ty1 = cvFloor(tyf);
int ty2 = ty1 + 1;
// 差值作为插值的比例
const float ya = tyf - ty1;
ty1 = std::max(ty1, 0);
ty2 = std::min(ty2, tilesY_ - 1);
const uchar* lutPlane1 = lut_.ptr(ty1 * tilesX_); // 当前块的直方图
const uchar* lutPlane2 = lut_.ptr(ty2 * tilesX_); // 向下一块的直方图
for (int x = 0; x < src_.cols; ++x)
{
const float txf = (static_cast<float>(x) / tileSize_.width) - 0.5f;
int tx1 = cvFloor(txf);
int tx2 = tx1 + 1;
// 差值作为插值的比例
const float xa = txf - tx1;
tx1 = std::max(tx1, 0);
tx2 = std::min(tx2, tilesX_ - 1);
// src_.ptr<uchar>(y)[x]
const int srcVal = srcRow[x];
// 索引 LUT
const size_t ind1 = tx1 * lut_step + srcVal;
const size_t ind2 = tx2 * lut_step + srcVal; // 向右一块的直方图
float res = 0;
// 根据直方图的值进行插值
// lut_.ptr(ty1 * tilesX_)[tx1 * lut_step + srcVa] => lut_[ty1][tx1][srcVal]
res += lutPlane1[ind1] * ((1.0f - xa) * (1.0f - ya));
res += lutPlane1[ind2] * ((xa) * (1.0f - ya));
res += lutPlane2[ind1] * ((1.0f - xa) * (ya));
res += lutPlane2[ind2] * ((xa) * (ya));
dstRow[x] = cv::saturate_cast<uchar>(res);
}
}
}
class CLAHE_Impl : public cv::CLAHE
{
public:
CLAHE_Impl(double clipLimit = 40.0, int tilesX = 8, int tilesY = 8);
cv::AlgorithmInfo* info() const; // Algorithm类工厂方法封装相关
void apply(cv::InputArray src, cv::OutputArray dst);
void setClipLimit(double clipLimit);
double getClipLimit() const;
void setTilesGridSize(cv::Size tileGridSize);
cv::Size getTilesGridSize() const;
void collectGarbage();
private:
double clipLimit_;
int tilesX_;
int tilesY_;
cv::Mat srcExt_;
cv::Mat lut_;
};
CLAHE_Impl::CLAHE_Impl(double clipLimit, int tilesX, int tilesY) :
clipLimit_(clipLimit), tilesX_(tilesX), tilesY_(tilesY)
{
}
// Algorithm类工厂方法封装相关
//CV_INIT_ALGORITHM(CLAHE_Impl, "CLAHE",
// obj.info()->addParam(obj, "clipLimit", obj.clipLimit_);
//obj.info()->addParam(obj, "tilesX", obj.tilesX_);
//obj.info()->addParam(obj, "tilesY", obj.tilesY_))
void CLAHE_Impl::apply(cv::InputArray _src, cv::OutputArray _dst)
{
cv::Mat src = _src.getMat();
CV_Assert(src.type() == CV_8UC1);
_dst.create(src.size(), src.type());
cv::Mat dst = _dst.getMat();
const int histSize = 256;
// 准备 LUT,tilesX_*tilesY_个块,每个块都有256个柱子的直方图
lut_.create(tilesX_ * tilesY_, histSize, CV_8UC1);
cv::Size tileSize;
cv::Mat srcForLut;
// 如果分块刚好(整除)
if (src.cols % tilesX_ == 0 && src.rows % tilesY_ == 0)
{
tileSize = cv::Size(src.cols / tilesX_, src.rows / tilesY_);
srcForLut = src;
}
// 否则对原图进行扩充
else
{
cv::copyMakeBorder(src, srcExt_, 0, tilesY_ - (src.rows % tilesY_), 0, tilesX_ - (src.cols % tilesX_), cv::BORDER_REFLECT_101);
tileSize = cv::Size(srcExt_.cols / tilesX_, srcExt_.rows / tilesY_);
srcForLut = srcExt_;
}
const int tileSizeTotal = tileSize.area();
const float lutScale = static_cast<float>(histSize - 1) / tileSizeTotal; // △
// 计算实际的clipLimit
int clipLimit = 0;
if (clipLimit_ > 0.0)
{
clipLimit = static_cast<int>(clipLimit_ * tileSizeTotal / histSize);
clipLimit = std::max(clipLimit, 1);
}
// 分块并行计算: LUT
CLAHE_CalcLut_Body calcLutBody(srcForLut, lut_, tileSize, tilesX_, clipLimit, lutScale);
cv::parallel_for_(cv::Range(0, tilesX_ * tilesY_), calcLutBody);
// 分块并行计算: 根据直方图插值
CLAHE_Interpolation_Body interpolationBody(src, dst, lut_, tileSize, tilesX_, tilesY_);
cv::parallel_for_(cv::Range(0, src.rows), interpolationBody);
}
void CLAHE_Impl::setClipLimit(double clipLimit)
{
clipLimit_ = clipLimit;
}
double CLAHE_Impl::getClipLimit() const
{
return clipLimit_;
}
void CLAHE_Impl::setTilesGridSize(cv::Size tileGridSize)
{
tilesX_ = tileGridSize.width;
tilesY_ = tileGridSize.height;
}
cv::Size CLAHE_Impl::getTilesGridSize() const
{
return cv::Size(tilesX_, tilesY_);
}
void CLAHE_Impl::collectGarbage()
{
srcExt_.release();
lut_.release();
}
}
cv::Ptr<cv::CLAHE> cv::createCLAHE(double clipLimit, cv::Size tileGridSize)
{
return new CLAHE_Impl(clipLimit, tileGridSize.width, tileGridSize.height);
}
main.cpp
int main(int argc, char** argv)
{
cv::Mat inp_img = cv::imread("D:/Pictures/beard.jpg");
if (!inp_img.data) {
cout << "Something Wrong";
return -1;
}
namedWindow("Input Image", CV_WINDOW_AUTOSIZE);
cv::imshow("Input Image", inp_img);
cv::Mat clahe_img;
cv::cvtColor(inp_img, clahe_img, CV_BGR2Lab);
std::vector<cv::Mat> channels(3);
cv::split(clahe_img, channels);
cv::Ptr<cv::CLAHE> clahe = cv::createCLAHE();
// 直方图的柱子高度大于计算后的ClipLimit的部分被裁剪掉,然后将其平均分配给整张直方图
// 从而提升整个图像
clahe->setClipLimit(4.); // (int)(4.*(8*8)/256)
//clahe->setTilesGridSize(Size(8, 8)); // 将图像分为8*8块
cv::Mat dst;
clahe->apply(channels[0], dst);
dst.copyTo(channels[0]);
cv::merge(channels, clahe_img);
cv::Mat image_clahe;
cv::cvtColor(clahe_img, image_clahe, CV_Lab2BGR);
//cout << cvFloor(-1.5) << endl;
namedWindow("CLAHE Image", CV_WINDOW_AUTOSIZE);
cv::imshow("CLAHE Image", image_clahe);
imwrite("out.jpg", image_clahe);
cv::waitKey(0);
destroyAllWindows();
return 0;
}
注意:cv::ParallelLoopBody 位于 %OpenCV%\opencv\sources\modules\core\src\parallel.cpp
延伸阅读:Algorithm类工厂方法封装相关