从编码图像和视频中提取 DCT 系数

Question

有没有一种方法可以轻松地从编码图像和视频中提取 DCT 系数（和量化参数）？任何解码器软件都必须使用它们来解码块 DCT 编码的图像和视频。所以我很确定解码器知道它们是什么。有没有办法将它们暴露给使用解码器的人？

我正在实现一些直接在 DCT 域中工作的视频质量评估算法。目前，我的大部分代码都使用 OpenCV，因此如果有人知道使用该框架的解决方案，那就太好了。我不介意使用其他库（也许是 libjpeg，但这似乎仅适用于静态图像），但我主要关心的是尽可能少地执行特定于格式的工作（我不想重新发明轮子并编写我自己的解码器）。我希望能够打开 OpenCV 可以打开的任何视频/图像（H.264、MPEG、JPEG 等），并且如果它是块 DCT 编码的，以获得 DCT 系数。

在最坏的情况下，我知道我可以编写自己的块 DCT 代码，通过它运行解压缩的帧/图像，然后我会回到 DCT 域。这并不是一个优雅的解决方案，我希望我能做得更好。

目前，我使用相当常见的 OpenCV 样板来打开图像：

IplImage *image = cvLoadImage(filename);
// Run quality assessment metric

我用于视频的代码同样简单：

CvCapture *capture = cvCaptureFromAVI(filename);    
while (cvGrabFrame(capture))
{
    IplImage *frame = cvRetrieveFrame(capture);
    // Run quality assessment metric on frame
}
cvReleaseCapture(&capture);

在这两种情况下，我都会得到 BGR 格式的 3 通道 IplImage。有什么方法可以得到 DCT 系数吗？

Answer 1

好吧，我读了一些书，我原来的问题似乎是一厢情愿的例子。

基本上，不可能从 H.264 视频帧中获取 DCT 系数，原因很简单，H.264 不使用 DCT。它使用不同的变换（整数变换）。接下来，该变换的系数不一定会逐帧变化 - H.264 更智能，因为它将帧分割成切片。应该可以通过特殊的解码器获得这些系数，但我怀疑 OpenCV 是否会向用户公开它。

对于 JPEG 来说，情况要乐观一些。正如我怀疑的那样， libjpeg 为您公开了 DCT 系数。我编写了一个小应用程序来证明它可以工作（源代码在最后）。它使用每个块的 DC 项生成一个新图像。由于 DC 项等于块平均值（经过适当缩放后），因此 DC 图像是输入 JPEG 图像的下采样版本。

编辑：固定源图像中的缩放比例

原始图像 (512 x 512)：

DC 图像 ( 64x64): 亮度 Cr Cb RGB

源 (C++)：

#include <stdio.h>
#include <assert.h>

#include <cv.h>    
#include <highgui.h>

extern "C"
{
#include "jpeglib.h"
#include <setjmp.h>
}

#define DEBUG 0
#define OUTPUT_IMAGES 1

/*
 * Extract the DC terms from the specified component.
 */
IplImage *
extract_dc(j_decompress_ptr cinfo, jvirt_barray_ptr *coeffs, int ci)
{
    jpeg_component_info *ci_ptr = &cinfo->comp_info[ci];
    CvSize size = cvSize(ci_ptr->width_in_blocks, ci_ptr->height_in_blocks);
    IplImage *dc = cvCreateImage(size, IPL_DEPTH_8U, 1);
    assert(dc != NULL);

    JQUANT_TBL *tbl = ci_ptr->quant_table;
    UINT16 dc_quant = tbl->quantval[0];

#if DEBUG
    printf("DCT method: %x\n", cinfo->dct_method);
    printf
    (
        "component: %d (%d x %d blocks) sampling: (%d x %d)\n", 
        ci, 
        ci_ptr->width_in_blocks, 
        ci_ptr->height_in_blocks,
        ci_ptr->h_samp_factor, 
        ci_ptr->v_samp_factor
    );

    printf("quantization table: %d\n", ci);
    for (int i = 0; i < DCTSIZE2; ++i)
    {
        printf("% 4d ", (int)(tbl->quantval[i]));
        if ((i + 1) % 8 == 0)
            printf("\n");
    }

    printf("raw DC coefficients:\n");
#endif

    JBLOCKARRAY buf =
    (cinfo->mem->access_virt_barray)
    (
        (j_common_ptr)cinfo,
        coeffs[ci],
        0,
        ci_ptr->v_samp_factor,
        FALSE
    );
    for (int sf = 0; (JDIMENSION)sf < ci_ptr->height_in_blocks; ++sf)
    {
        for (JDIMENSION b = 0; b < ci_ptr->width_in_blocks; ++b)
        {
            int intensity = 0;

            intensity = buf[sf][b][0]*dc_quant/DCTSIZE + 128;
            intensity = MAX(0,   intensity);
            intensity = MIN(255, intensity);

            cvSet2D(dc, sf, (int)b, cvScalar(intensity));

#if DEBUG
            printf("% 2d ", buf[sf][b][0]);                        
#endif
        }
#if DEBUG
        printf("\n");
#endif
    }

    return dc;

}

IplImage *upscale_chroma(IplImage *quarter, CvSize full_size)
{
    IplImage *full = cvCreateImage(full_size, IPL_DEPTH_8U, 1);
    cvResize(quarter, full, CV_INTER_NN);
    return full;
}

GLOBAL(int)
read_JPEG_file (char * filename, IplImage **dc)
{
  /* This struct contains the JPEG decompression parameters and pointers to
   * working space (which is allocated as needed by the JPEG library).
   */
  struct jpeg_decompress_struct cinfo;

  struct jpeg_error_mgr jerr;
  /* More stuff */
  FILE * infile;        /* source file */

  /* In this example we want to open the input file before doing anything else,
   * so that the setjmp() error recovery below can assume the file is open.
   * VERY IMPORTANT: use "b" option to fopen() if you are on a machine that
   * requires it in order to read binary files.
   */

  if ((infile = fopen(filename, "rb")) == NULL) {
    fprintf(stderr, "can't open %s\n", filename);
    return 0;
  }

  /* Step 1: allocate and initialize JPEG decompression object */

  cinfo.err = jpeg_std_error(&jerr);

  /* Now we can initialize the JPEG decompression object. */
  jpeg_create_decompress(&cinfo);

  /* Step 2: specify data source (eg, a file) */

  jpeg_stdio_src(&cinfo, infile);

  /* Step 3: read file parameters with jpeg_read_header() */

  (void) jpeg_read_header(&cinfo, TRUE);
  /* We can ignore the return value from jpeg_read_header since
   *   (a) suspension is not possible with the stdio data source, and
   *   (b) we passed TRUE to reject a tables-only JPEG file as an error.
   * See libjpeg.txt for more info.
   */

  /* Step 4: set parameters for decompression */

  /* In this example, we don't need to change any of the defaults set by
   * jpeg_read_header(), so we do nothing here.
   */

  jvirt_barray_ptr *coeffs = jpeg_read_coefficients(&cinfo);

  IplImage *y    = extract_dc(&cinfo, coeffs, 0);
  IplImage *cb_q = extract_dc(&cinfo, coeffs, 1);
  IplImage *cr_q = extract_dc(&cinfo, coeffs, 2);

  IplImage *cb = upscale_chroma(cb_q, cvGetSize(y));
  IplImage *cr = upscale_chroma(cr_q, cvGetSize(y));

  cvReleaseImage(&cb_q);
  cvReleaseImage(&cr_q);

#if OUTPUT_IMAGES
  cvSaveImage("y.png",   y);
  cvSaveImage("cb.png", cb);
  cvSaveImage("cr.png", cr);
#endif

  *dc = cvCreateImage(cvGetSize(y), IPL_DEPTH_8U, 3);
  assert(dc != NULL);

  cvMerge(y, cr, cb, NULL, *dc);

  cvReleaseImage(&y);
  cvReleaseImage(&cb);
  cvReleaseImage(&cr);

  /* Step 7: Finish decompression */

  (void) jpeg_finish_decompress(&cinfo);
  /* We can ignore the return value since suspension is not possible
   * with the stdio data source.
   */

  /* Step 8: Release JPEG decompression object */

  /* This is an important step since it will release a good deal of memory. */
  jpeg_destroy_decompress(&cinfo);

  fclose(infile);

  return 1;
}

int 
main(int argc, char **argv)
{
    int ret = 0;
    if (argc != 2)
    {
        fprintf(stderr, "usage: %s filename.jpg\n", argv[0]);
        return 1;
    }
    IplImage *dc = NULL;
    ret = read_JPEG_file(argv[1], &dc);
    assert(dc != NULL);

    IplImage *rgb = cvCreateImage(cvGetSize(dc), IPL_DEPTH_8U, 3);
    cvCvtColor(dc, rgb, CV_YCrCb2RGB);

#if OUTPUT_IMAGES
    cvSaveImage("rgb.png", rgb);
#else
    cvNamedWindow("DC", CV_WINDOW_AUTOSIZE); 
    cvShowImage("DC", rgb);
    cvWaitKey(0);
#endif

    cvReleaseImage(&dc);
    cvReleaseImage(&rgb);

    return 0;
}

Answer 2

您可以使用libjpeg来提取jpeg文件的dct数据，但是对于h.264视频文件，我找不到任何为您提供dct数据的开源代码（实际上是整数DCT数据）。但您可以使用 h.264 开源软件，例如 JM、JSVM 或 x264。在这两个源文件中，您必须找到它们使用dct函数的特定函数，并将其更改为您想要的形式，以获得输出dct数据。

对于图像：
使用以下代码，在 read_jpeg_file( infilename, v, quant_tbl ) 之后，v 和 quant_tbl 将获得 dct 数据 和 jpeg 图像的量化表。

我使用 Qvector 来存储我的输出数据，将其更改为您首选的 C++ 数组列表。

---

#include <iostream>
#include <stdio.h>
#include <jpeglib.h>
#include <stdlib.h>
#include <setjmp.h>
#include <fstream>

#include <QVector>

int read_jpeg_file( char *filename, QVector<QVector<int> > &dct_coeff, QVector<unsigned short> &quant_tbl)
{
    struct jpeg_decompress_struct cinfo;
    struct jpeg_error_mgr jerr;
    FILE * infile;

    if ((infile = fopen(filename, "rb")) == NULL) {
      fprintf(stderr, "can't open %s\n", filename);
      return 0;
    }

    cinfo.err = jpeg_std_error(&jerr);
    jpeg_create_decompress(&cinfo);
    jpeg_stdio_src(&cinfo, infile);
    (void) jpeg_read_header(&cinfo, TRUE);

    jvirt_barray_ptr *coeffs_array = jpeg_read_coefficients(&cinfo);
    for (int ci = 0; ci < 1; ci++)
    {
        JBLOCKARRAY buffer_one;
        JCOEFPTR blockptr_one;
        jpeg_component_info* compptr_one;
        compptr_one = cinfo.comp_info + ci;

        for (int by = 0; by < compptr_one->height_in_blocks; by++)
        {
            buffer_one = (cinfo.mem->access_virt_barray)((j_common_ptr)&cinfo, coeffs_array[ci], by, (JDIMENSION)1, FALSE);
            for (int bx = 0; bx < compptr_one->width_in_blocks; bx++)
            {
                blockptr_one = buffer_one[0][bx];
                QVector<int> tmp;
                for (int bi = 0; bi < 64; bi++)
                {
                    tmp.append(blockptr_one[bi]);
                }
                dct_coeff.push_back(tmp);
            }
        }
    }


    // coantization table
    j_decompress_ptr dec_cinfo  = (j_decompress_ptr) &cinfo;
    jpeg_component_info *ci_ptr = &dec_cinfo->comp_info[0];
    JQUANT_TBL *tbl = ci_ptr->quant_table;

    for(int ci =0 ; ci < 64; ci++){
        quant_tbl.append(tbl->quantval[ci]);
    }

    return 1;
}

int main()
{
    QVector<QVector<int> > v;
    QVector<unsigned short> quant_tbl;
    char *infilename = "your_image.jpg";

    std::ofstream out;
    out.open("out_dct.txt");


    if( read_jpeg_file( infilename, v, quant_tbl ) > 0 ){

        for(int j = 0; j < v.size(); j++ ){
                for (int i = 0; i < v[0].size(); ++i){
                    out << v[j][i] << "\t";
            }
            out << "---------------" << std::endl;
        }

        out << "\n\n\n" << std::string(10,'-') << std::endl;
        out << "\nQauntization Table:" << std::endl;
        for(int i = 0; i < quant_tbl.size(); i++ ){
            out << quant_tbl[i] << "\t";
        }
    }
    else{
        std::cout << "Can not read, Returned With Error";
        return -1;
    }

    out.close();

return 0;
}