Recherche avancée

Sur d’autres sites (5395)

• avcodec/sanm : ignore codec45

  14 mars, par Manuel Lauss

  avcodec/sanm : ignore codec45
  Codec45 is used in some RA2 videos on top of codec37.
  It consists of 2 tables (768 and 32768 bytes), and datapackets
  
  like codec2 (delta-x, delta-y) with a pixel counter instead of a color value.
  
  It then reads the 4 surrounding pixels, looks up 3 separate values
  
  for each in table1, adds them together to form an index into
  
  table2 for a new pixel value, in a row.
  The data coming in gets the x/y coordinates out of the visible
  
  area very quickly (2-3 iterations) ; I don't see any visual
  
  difference between the ffmpeg-produced frames and ones captured
  
  with dosbox from rebel assault 2, which leads me to believe this
  
  codec never worked as intended.
  Signed-off-by : Manuel Lauss <manuel.lauss@gmail.com>
  

  • [DH] libavcodec/sanm.c

• FFMPEG spoils the extreme pixels when converting and cropping images [closed]

  20 décembre 2024, par Jacob

  I have a side image processing task in my work.
I am using the following command to translate a spherical panorama into a cubemap :

  


  Path/ffmpeg.exe -i "Path/PanoImageInput.png" -vf v360=equirect:c6x1:fin_pad=64:out_frot=002200,scale=-1:2048:flags=neighbor "Path/CubemapOutput.png"

  


  And then I slice it into separate images, 1 face for example :

  


  Path/ffmpeg.exe -i "Path/CubemapOutput.png" -vf crop=2048:2048:0:0 -sws_flags neighbor "Path/face_1.png"

  


  I get artifacts on both cubemap and images of this kind that really bother me :

  


  Cubemap image junction
The junction and part of the edge of the cubemap image

  


  The pixels at the junctions of the 6 images in cubemap and the outermost pixels of any images change their color slightly. This eventually leads to the appearance of visible seams in the scene assembled from the images.

  


  Is there any way I can get rid of them ?

  


  I have tried different interpolation methods in both functions : fast, bicubric, gauss, etc. It doesn't seem to have any effect.

  


  I also tried to crop a couple of pixels less, something like : crop=2040:2040:4:4

  


  I thought it was all because of cubemap and its distortions. Anyway, at the edges of the image, the extreme pixels change their hue.

  


  I also hoped that with the help of pad, during the conversion, I could somehow control the area where the pixels deteriorate, but it absolutely does not matter what I put there - fin_pad, fout_pad and any numbers for them, remove scale, set scale - neither the final image size changes, nor any extra fields appear. Most likely, I just don't understand what this function is really supposed to mean.

  


• Output video like Instagram stories with ffmpeg

  1er novembre 2024, par arman naeimi

  I'm trying to build a small component like Instagram stories with React & Vite & Typescript. The user can zoom in, zoom out and rotate their videos easily on the screen with a touch.

  


  Using ffmpeg I want to get an output like what user sees in his browser. definitely what the user sees on the screen of his phone and the area he is working on is smaller and our output is a video with larger dimensions in the size of 1080 x 1920. So I adjusted the values before sending them to server.

  


  Front-end Code :

  


    const handleUpload = async () => {
    if (isVideo && video && canvasRef.current && mediaDimensions) {
      const TARGET_WIDTH = 1080; // Target output width
      const TARGET_HEIGHT = 1920; // Target output height

      const scaleX = TARGET_WIDTH / mediaDimensions.width;
      const scaleY = TARGET_HEIGHT / mediaDimensions.height;

      const adjustedScale = Math.min(scaleX, scaleY) * scale;

      const adjustedRotate = rotation * (180 / Math.PI);

      const adjustedOffsetX =
        ((offset.x * mediaDimensions.width) / canvasRef.current.width) *
        adjustedScale;
      const adjustedOffsetY =
        ((offset.y * mediaDimensions.height) / canvasRef.current.height) *
        adjustedScale;

      const formData = new FormData();
      formData.append('video', video);
      formData.append(
        'transform',
        JSON.stringify({
          scale: adjustedScale,
          rotate: adjustedRotate,
          offsetX: adjustedOffsetX,
          offsetY: adjustedOffsetY,
        })
      );

      const response = await axios.post('/api/watermark', formData, {
        responseType: 'blob',
      });

      const url = URL.createObjectURL(new Blob([response.data]));
      setVideoUrl(url);
    }
  };


  


  The ffmpeg command that I wrote :

  


    const command = `
  ffmpeg -i ${file.path} -vf "
  scale='min(1080/${transform.scale}, iw*${transform.scale}):min(1920/${transform.scale}, ih*${transform.scale})',
  fps=30,
  format=yuv420p,
  rotate=${transform.rotate}*PI/180:ow=rotw(${transform.rotate}*PI/180):oh=roth(${transform.rotate}*PI/180),
  pad=1080:1920:(1080-iw*${transform.scale})/2+${transform.offsetX}:(1920-ih*${transform.scale})/2+${transform.offsetY},
  crop=1080:1920:(max((iw-1080)/2\\,0) + ${transform.offsetX}):(max((ih-1920)/2\\,0) + ${transform.offsetY})
  " -c:v libx264 -preset fast -crf 28 -y ${outputPath}
`;;


  


  I wrote different commands but did not get the desired output and kept getting errors. I want a command that behaves exactly like Instagram. If the user zooms the video out of the canvas, the output is still 1080x1920 and only outputs as much as is visible. If it was smaller, its place is only adjusted.