www.pudn.com > 99273898StereoMatch_1_0.zip > Histogram1D.cpp

/////////////////////////////////////////////////////////////////////////// 
// 
// NAME 
//  Histogram1D.cpp -- compute a 1-D histogram 
// 
// SPECIFICATION 
//  int  Histogram1D(CImageOf img, int n_bands, CByteImage mask, 
//                   std::vector& counts, 
//                   float& min_val, float& max_val, float& step_size, int& n_bins); 
//  void Histogram1D(CImageOf img, int n_bands, CByteImage mask, 
//                   CByteImage& figure, 
//                   float& min_val, float& max_val, float& step_size, int& n_bins, 
//                   int& height, float& v_scale); 
// 
// SEE ALSO 
//  Histogram1D.h       longer description 
//  Image.h             image class definition 
// 
// Copyright © Richard Szeliski, 2001. 
// See Copyright.h for more details 
// 
/////////////////////////////////////////////////////////////////////////// 
 
#include "Image.h" 
#include "Error.h" 
#include  
#include  
#include  
#include  
#include "Histogram1D.h" 
 
template  
extern int  Histogram1D(CImageOf img, int n_bands, CByteImage mask, 
                        std::vector& counts, 
                        float& min_val, float& max_val, float& step_size, int& n_bins) 
{ 
    // Set up the dimensions 
    CShape sh = img.Shape(), msh = mask.Shape(); 
    bool valid_mask = (sh.width == msh.width) && (sh.height == msh.height); 
    n_bands = (n_bands > 0) ? n_bands : sh.nBands; 
 
    // Compute the min and max values 
    if (min_val >= max_val) 
    { 
        T min_v = img.MaxVal(); 
        T max_v = img.MinVal(); 
        for (int y = 0; y < sh.height; y++) 
        { 
            T*     p = &img.Pixel(0, y, 0); 
            uchar* m = (valid_mask) ? &mask.Pixel(0, y, 0) : 0; 
            for (int x = 0; x < sh.width; x++) 
            { 
                if (m && m[x] == 0) 
                    break; 
                for (int b = 0; b < n_bands; b++, p++) 
                { 
                    min_v = __min(min_v, p[0]); 
                    max_v = __max(max_v, p[0]); 
                } 
            } 
        } 
        max_val = max_v; 
        min_val = min_v; 
    } 
 
    // Compute the step size and number of bins 
    if (n_bins <= 0 && step_size > 0.0) 
        n_bins = int(ceil((max_val - min_val) / step_size)); 
    else if (step_size <= 0 && n_bins > 0) 
        step_size = (max_val - min_val) / (float) n_bins; 
    else if (n_bins <= 0 && step_size <= 0) 
        throw CError("Histogram1D: both step_size and n_bins can't be 0"); 
    assert(n_bins > 0); 
 
    // Fill in the bins 
    counts.resize(n_bins); 
    std::fill_n(counts.begin(), n_bins, 0); 
    float inv_step_size = 1.0f / step_size; 
    float min_v = min_val; 
    for (int y = 0; y < sh.height; y++) 
    { 
        T*     p = &img.Pixel(0, y, 0); 
        uchar* m = (valid_mask) ? &mask.Pixel(0, y, 0) : 0; 
        for (int x = 0; x < sh.width; x++) 
        { 
            if (m && m[x] == 0) 
                break; 
            for (int b = 0; b < n_bands; b++, p++) 
            { 
                float val = p[0]; 
                int bin = int((val - min_v) * inv_step_size); 
                bin = __min(__max(0, bin), n_bins-1); 
                counts[bin] += 1; 
            } 
        } 
    } 
 
    // Return the largest bin size 
    int max_count = (std::max_element(counts.begin(), counts.end()))[0]; 
    return max_count; 
} 
 
template  
extern void Histogram1D(CImageOf img, int n_bands, CByteImage mask, 
                        CByteImage& figure, 
                        float& min_val, float& max_val, float& step_size, int& n_bins, 
                        int& height, float& v_scale) 
{ 
    // Compute the histogram counts 
    std::vector counts; 
    int max_count = 
        Histogram1D(img, n_bands, mask, counts, 
                    min_val, max_val, step_size, n_bins); 
 
    // Compute the height scaling 
    if (height <= 0 && v_scale > 0.0) 
        height = int(ceil(max_count * v_scale)); 
    else if (v_scale <= 0 && height > 0) 
        v_scale = height / (float) max_count; 
    else if (height <= 0 && step_size <= 0.0) 
        throw CError("Histogram1D: both height and v_scale can't be 0"); 
    assert(height > 0); 
 
    // Re-normalize the counts to be in the range [0,height) 
    for (int i = 0; i < n_bins; i++) 
        counts[i] = __min(height-1, int(counts[i] * v_scale + 0.5)); 
 
    // Draw the figure 
    CShape sh(n_bins, height, 1); 
    figure.ReAllocate(sh); 
    figure.ClearPixels(); 
    for (int y = 0; y < height; y++) 
    { 
        uchar* p = &figure.Pixel(0, y, 0); 
        int y2 = height-1 - y; 
        for (int x = 0; x < n_bins; x++) 
            p[x] = (y2 >= counts[x]) ? 255 : 0; 
    } 
} 
 
extern void Histogram1DInstantiate() 
{ 
    CByteImage mask, figure; 
    int n_bins, height; 
    float mn, mx, st, v_scale; 
    Histogram1D(CByteImage(),  0, mask, figure, mn, mx, st, n_bins, height, v_scale); 
    Histogram1D(CIntImage(),   0, mask, figure, mn, mx, st, n_bins, height, v_scale); 
    Histogram1D(CFloatImage(), 0, mask, figure, mn, mx, st, n_bins, height, v_scale); 
}