【特征提取】MultiBlock-LBP特征

MB-LBP特征，注意区分是Multi-block LBP还是MultiScale Block LBP,区别是是否使用了多尺度方法。其中Multiscale Block LBP，来源于论文，中科院的人发明的，opencv源码在Traincascade级联目标训练检测中的LBP特征使用的就是MultiScale Block LBP。

generateFeatures

void CvLBPEvaluator::generateFeatures()
{
    int offset = winSize.width + ;
    for( int x = ; x < winSize.width; x++ )
        for( int y = ; y < winSize.height; y++ )
            for( int w = ; w <= winSize.width / ; w++ )
                for( int h = ; h <= winSize.height / ; h++ )
                    if ( (x+*w <= winSize.width) && (y+*h <= winSize.height) )
                        features.push_back( Feature(offset, x, y, w, h ) );
    numFeatures = (int)features.size();
}

Feature

CvLBPEvaluator::Feature::Feature( int offset, int x, int y, int _blockWidth, int _blockHeight )
{
    Rect tr = rect = cvRect(x, y, _blockWidth, _blockHeight);
    CV_SUM_OFFSETS( p[], p[], p[], p[], tr, offset )
    tr.x += *rect.width;
    CV_SUM_OFFSETS( p[], p[], p[], p[], tr, offset )
    tr.y +=*rect.height;
    CV_SUM_OFFSETS( p[], p[], p[], p[], tr, offset )
    tr.x -= *rect.width;
    CV_SUM_OFFSETS( p[], p[], p[], p[], tr, offset )
}

calc

inline uchar CvLBPEvaluator::Feature::calc(const cv::Mat &_sum, size_t y) const
{
    const int* psum = _sum.ptr<int>((int)y);
    int cval = psum[p[]] - psum[p[]] - psum[p[]] + psum[p[]];

    return (uchar)((psum[p[]] - psum[p[]] - psum[p[]] + psum[p[]] >= cval ?  : ) |   //
        (psum[p[]] - psum[p[]] - psum[p[]] + psum[p[]] >= cval ?  : ) |    //
        (psum[p[]] - psum[p[]] - psum[p[]] + psum[p[]] >= cval ?  : ) |    //
        (psum[p[]] - psum[p[]] - psum[p[]] + psum[p[]] >= cval ?  : ) |  //
        (psum[p[]] - psum[p[]] - psum[p[]] + psum[p[]] >= cval ?  : ) | //
        (psum[p[]] - psum[p[]] - psum[p[]] + psum[p[]] >= cval ?  : ) |  //
        (psum[p[]] - psum[p[]] - psum[p[]] + psum[p[]] >= cval ?  : ) |   //
        (psum[p[]] - psum[p[]] - psum[p[]] + psum[p[]] >= cval ?  : ));     //
}

CV_SUM_OFFSETS

#define CV_SUM_OFFSETS( p0, p1, p2, p3, rect, step )                      \
    /* (x, y) */                                                          \
    (p0) = (rect).x + (step) * (rect).y;                                  \
    /* (x + w, y) */                                                      \
    (p1) = (rect).x + (rect).width + (step) * (rect).y;                   \
    /* (x, y + h) */                                    \
    (p2) = (rect).x + (step) * ((rect).y + (rect).height);                \
    /* (x + w, y + h) */                                                  \
    (p3) = (rect).x + (rect).width + (step) * ((rect).y + (rect).height);

参考

1. LBP特征原理及代码实现；

2. opencv源码解析之LBP特征；

3. 机器学习之LBP特征；

4. matalb code之MB-LBP特征；

5. Face Detection Based on Multi-Block LBP Representation;

6. Learning Multi-scale Block Local Binary Patterns for Face Recognition;

完