Recherche avancée

Sur d’autres sites (11427)

• How to extract good quality frame in pixel format from live mpeg-ts stream ?

  2 février 2024, par Aven

  I use the command below to stream my ts file for my programe :

  


  ffmpeg -re -i independent_frame.ts -f mpegts udp://127.0.0.1:10002


  


  Below is the code of my programe, I'm trying to extract each frame in pixel format from the stream by using FFmpeg, but I found the quality is not good as it should be of the extracted frame ( I know OpenCV can do this, But I have other reason have to use FFmpeg ). I tried to put -qscale:v 2 in FFmpeg command, but it seems change nothing. Does anyone know how to enhance the frame quality in this case ? Thanks !

  


  stream = 'udp://127.0.0.1:10002'
H, W = 720, 1280

command = [ 'ffmpeg',
            '-i', stream ,
            '-pix_fmt', 'bgr24', # brg24 for matching OpenCV
            '-qscale:v', '2', 
            '-f', 'rawvideo',
            'pipe:' ]

# Execute FFmpeg as sub-process with stdout as a pipe
process = subprocess.Popen(command, stdout=subprocess.PIPE, bufsize=10**8)#10**8

# Load individual frames in a loop
nb_img = H*W*3

# Read decoded video frames from the PIPE until no more frames to read
num = 0
while True:
    # Read decoded video frame (in raw video format) from stdout process.
    buffer = process.stdout.read(W*H*3)

    # Break the loop if buffer length is not W*H*3 (when FFmpeg streaming ends).
    if len(buffer) != W*H*3:
        break

    img = np.frombuffer(buffer, np.uint8).reshape(H, W, 3)
    num +=1

    if num == 435:
        cv2.imwrite('435.png', img)
        cv2.imshow('img', img)  # Show the image for testing
        cv2.waitKey(0)
        cv2.destroyAllWindows()
        break

process.stdout.close()
process.wait()
cv2.destroyAllWindows()


  


• How to parse WMV (ASF) file ? Can't find length of data packets

  9 mars 2018, par bukkojot

  I try to parse WMV (ASF) files without any SDK, just by decoding raw bytes. Now I have problem with ASF_Data_Object, where I can’t find length of data packet. More precise, Single payload data packet.

  See image :
  

  Here I have 9 packets, but unable to find size of individual packet. How I can determine border between packets ?

  I think, my problem at byte 0x411, where field "Length type flags". As you can see, here 0 value, so all flags are zero. Even Packet Length Type.

  
  

  Yes, 0 value here allowed here. But how to read this type of content ?

  This is now compressed payload, as replication data is 8, not 1. So, this is single payload without additional fields of size.

  Sample of WMV file : https://files.catbox.moe/b51l2j.wmv

• aarch64 : vp9itxfm : Skip empty slices in the first pass of idct_idct 16x16 and 32x32

  9 janvier 2017, par Martin Storsjö

  aarch64 : vp9itxfm : Skip empty slices in the first pass of idct_idct 16x16 and 32x32
  This work is sponsored by, and copyright, Google.
  Previously all subpartitions except the eob=1 (DC) case ran with
  
  the same runtime :
  vp9_inv_dct_dct_16x16_sub16_add_neon :   1373.2
  
  vp9_inv_dct_dct_32x32_sub32_add_neon :   8089.0
  By skipping individual 8x16 or 8x32 pixel slices in the first pass,
  
  we reduce the runtime of these functions like this :
  vp9_inv_dct_dct_16x16_sub1_add_neon :     235.3
  
  vp9_inv_dct_dct_16x16_sub2_add_neon :    1036.7
  
  vp9_inv_dct_dct_16x16_sub4_add_neon :    1036.7
  
  vp9_inv_dct_dct_16x16_sub8_add_neon :    1036.7
  
  vp9_inv_dct_dct_16x16_sub12_add_neon :   1372.1
  
  vp9_inv_dct_dct_16x16_sub16_add_neon :   1372.1
  
  vp9_inv_dct_dct_32x32_sub1_add_neon :     555.1
  
  vp9_inv_dct_dct_32x32_sub2_add_neon :    5190.2
  
  vp9_inv_dct_dct_32x32_sub4_add_neon :    5180.0
  
  vp9_inv_dct_dct_32x32_sub8_add_neon :    5183.1
  
  vp9_inv_dct_dct_32x32_sub12_add_neon :   6161.5
  
  vp9_inv_dct_dct_32x32_sub16_add_neon :   6155.5
  
  vp9_inv_dct_dct_32x32_sub20_add_neon :   7136.3
  
  vp9_inv_dct_dct_32x32_sub24_add_neon :   7128.4
  
  vp9_inv_dct_dct_32x32_sub28_add_neon :   8098.9
  
  vp9_inv_dct_dct_32x32_sub32_add_neon :   8098.8
  I.e. in general a very minor overhead for the full subpartition case due
  
  to the additional cmps, but a significant speedup for the cases when we
  
  only need to process a small part of the actual input data.
  This is cherrypicked from libav commits
  
  cad42fadcd2c2ae1b3676bb398844a1f521a2d7b and
  
  a0c443a3980dc22eb02b067ac4cb9ffa2f9b04d2.
  Signed-off-by : Michael Niedermayer <michael@niedermayer.cc>
  

  • [DH] libavcodec/aarch64/vp9itxfm_neon.S