pHash 返回不同的哈希长度

Question

我有以下代码：

指纹.cpp：

#include <iostream>
#include <filesystem>

#include "ImageHash.h"

#include "opencv2/core.hpp"
#include "opencv2/highgui.hpp"
#include <opencv2/imgcodecs.hpp>

using namespace std;


void usage (char** argv) noexcept  {
    cout << "Usage: " << argv[0] << " <image or dir>" << endl << "Example: " << argv[0] << " template.png" << endl;
}

int main(int argc, char** argv) {
    if (argc != 2) {
        usage(argv);
        return 1;
    }

    auto imgHash = ImageHash();

    if (std::filesystem::is_directory(argv[1])) {
        for (const auto& dirEntry : std::filesystem::recursive_directory_iterator(argv[1])) {
            auto filePath = dirEntry.path();
            std::string fileName = dirEntry.path().filename().string();
        
            cv::Mat img = cv::imread(filePath.string(), cv::IMREAD_GRAYSCALE);
            img.resize(8*8);

            /*cv::imshow("img", img);
            cv::waitKey();*/
            
            cout << fileName << "\t\t" << imgHash.getHashString(img) << endl;
        }

    }
    else if (std::filesystem::is_regular_file(argv[1])){
        cv::Mat img = cv::imread(argv[1], cv::IMREAD_GRAYSCALE);
        cout << std::filesystem::path(argv[1]).filename().string() << "\t\t" << imgHash.getHashString(img) << endl;
    }
    else {
        usage(argv);
    }

    return 0;
}

ImageHash.cpp:

#include "ImageHash.h"

#include <iomanip> 

using namespace std;

ImageHash::ImageHash()
{
    pHash = cv::img_hash::PHash::create();
}

std::string ImageHash::convertHashToString(vector<bool> hash) {
    std::string ret = "";
    double h = 0;
    for (unsigned int i = 0; i < hash.size(); i++) {
        if (hash[i]) {
            h += pow(2, (i % 8));
        }

        if (i % 8 == 7) {
            std::stringstream buffer;
            buffer << std::hex << std::setfill('0') << std::setw(2) << h;
            ret += buffer.str();
            h = 0;
        }
    }
    return ret;
}

std::vector<bool> ImageHash::hex_str_to_hash(std::string inputString) {
    std::vector<bool> hash;
    size_t size = inputString.size() / 2;
    for (int i = 0; i < size; i++) {
        std::string str2 = inputString.substr(i * 2, 2);
        if (str2.empty()) {
            continue;
        }

        unsigned int value = 0;
        std::stringstream SS(str2);
        SS >> std::hex >> value;
        for (int j = 0; j < 8; j++) {
            bool check = !!((value >> j) & 1);
            hash.push_back(check);
        }
    }
    return hash;
}

std::vector<bool> ImageHash::matHashToBoolArr(cv::Mat const inHash) {
    const unsigned char* data = inHash.data;
    std::vector<bool> v;
    for (int i = 0; i < 8; i++) {
        unsigned char c = data[i];
        for (int j = 0; j < 8; j++) {
            int shift = (8 - j) - 1;
            bool val = ((c >> shift) & 1);
            v.push_back(val);
        }
    }
    return v;
}

std::vector<bool> ImageHash::computeHash(cv::Mat const input) {
    cv::Mat inHash;
    pHash->compute(input, inHash);
    return matHashToBoolArr(inHash);
}

std::string ImageHash::getHashString(cv::Mat const input) {
    return convertHashToString(computeHash(input));
}

//returns hamming distance
int ImageHash::getHashDistance(std::vector<bool>& hash1, std::vector<bool>& hash2) {
    //assert(hash1.size() == hash2.size());
    
    /* ToDo: I don't know why the size is not the same but if if is lets add padding to the smaller one
    This does not seem to make it work*/
    if (hash1.size() != hash2.size()) {
        auto smaller = hash1.size() < hash2.size() ? &hash1 : &hash2;
        auto larger = hash1.size() < hash2.size() ? hash2 : hash1;
        smaller->resize(larger.size());
        std::fill(begin(*smaller) + larger.size(), end(*smaller), 0);
    }


    int dist = 0;
    for (unsigned int i = 0; i < hash1.size(); i++) {
        dist += (hash1[i] != hash2[i]);
    }
    return dist;
}

由于某种原因，返回的哈希值根据输入图像的大小不同。我尝试在 getHashDistance() 中添加一些填充，但我认为我不应该这样做，因为现在当我计算两个相似图像之间的距离时，我得到的距离很大，就好像它们不相似一样。

你知道这是为什么吗？我正在将图像的大小调整为 8*8 img.resize(8*8);，虽然它会成功，但它不起作用。

谢谢 谢谢

Answer 1

auto imgHash = ImageHash();

会不会是这里的 auto 关键字导致运行时对返回类型（以及返回类型的大小）进行不同的推断？

Type Inference in C++ (auto and decltype)