Skip to content
Snippets Groups Projects

Full/Limited Video Levels incorrectly displayed

    • Closed Issue created by chris319 @chris319

    Version 3.0.11; Windows 10

    By actual measurement, luma levels of 235 are forced to 255 and luma levels of 16 are forced to 0. Input file is flagged as "full range". Unfortunately there is no way around this behavior.

    Please allow users to configure VLC to defeat this behavior. I had to revert to version 2.2.8 which works just fine, leaving 235 at 235 and 16 at 16. ffplay leaves 235 at 235 on the same file in both the player and waveform modes.

    Please see this exchange:

    https://forum.videolan.org/viewtopic.php?f=14&t=154616&p=507923&hilit=levels#p507923

    Sample files:

    http://www.chrisnology.info/video/Window_full.mp4

    The video levels of the full-range file are 0, 16, 235, 255.

    http://www.chrisnology.info/video/Window_limited.mp4

    Edited
    To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information

    Child items ...

    Activity

    • chris319
      chris319 @chris319 ·
      Author

      Update:

      Version 3.0.12

      Full range: video level 16 is forced to 0 (should be 16); video level 235 is forced to 255 (should be 235).

      Limited range: video level 16 is forced to 0 (should be 16); video level 235 forced to 255 (should be 235).

      VLC 3.0.12 is not displaying video levels accurately. On-screen video levels checked with eyedropper/color picker program.

      Video levels of test files have been checked using ffmpeg waveform. Test-signal files are correctly flagged as full/limited and checked with MediaInfo.

      Version 2.2.8 displays video levels correctly on both full- and limited-range videos.

    • Jean-Baptiste Kempf added Component::Video: Direct3D / Win32 label

      added Component::Video: Direct3D / Win32 label

    • Jean-Baptiste Kempf changed title from Full/Limited Video Levels to Full/Limited Video Levels incorrectly displayed

      changed title from Full/Limited Video Levels to Full/Limited Video Levels incorrectly displayed

    • Yafei Liu
      Yafei Liu @LiuYafei ·

      Big fan of VLC :)

      As I commented on this issue 21986, I think the problem is that matrix works in limited range applies to full range content, at least in my case. I don't have the latest 3rd party libraries to compile master branch on Windows, so I choose to verify my thought at this branch 4.0.0-dev(SHA1:1f2e534f). Full range content is transformed to limited range before color space matrix starts to work in HLSL shader

        float4 main( PS_INPUT In ) : SV_TARGET\
  {\
    float4 sample;\
    \
    %s /* sampling routine in sample */\
    sample = float4(fullRangeToStdio(sample.rgb), sample.a); /* full range to studio */\
    float4 rgba = mul(mul(sample, WhitePoint), Colorspace);\
    float opacity = rgba.a * Opacity;\
    float3 rgb = (float3)rgba;\
    rgb = sourceToLinear(rgb);\
    rgb = toneMapping(rgb);\
    rgb = adjustPeakLuminance(rgb);\
    rgb = linearToDisplay(rgb);\
    rgb = adjustRange(rgb);\
    return float4(rgb, saturate(opacity));\
  }\

      and change this piece of code a little bit

      int range_adjust = sys->display.colorspace->b_full_range;
    if (!IsRGBShader(format))
        range_adjust--; /* the YUV->RGB conversion already output full range */
    if (src_full_range)
        range_adjust--;

      And full range content looks right this time. I attach a picture, you can find these differences. Hope this could help :) diff

    • Jean-Baptiste Kempf marked this issue as related to #21986 (closed)

      marked this issue as related to #21986 (closed)

    • Jean-Baptiste Kempf marked this issue as related to #25651

      marked this issue as related to #25651

    • Jean-Baptiste Kempf marked this issue as related to #21594 (closed)

      marked this issue as related to #21594 (closed)

    • Jean-Baptiste Kempf marked this issue as related to #19723

      marked this issue as related to #19723

    • Jean-Baptiste Kempf
      Jean-Baptiste Kempf @jbk ·
      Owner

      Can you test VLC with D3D11, D3D9 and OpenGL outputs in 4.0?

    • Jean-Baptiste Kempf assigned to @robUx4

      assigned to @robUx4

    • Yafei Liu
      Yafei Liu @LiuYafei ·

      Sure!

      D3D11 and D3D9 still don't work correctly in 4.0 and the result looks exactly like nightly build version. OpenGL crashed in my build version, I think I didn't compile it correctly.

      1. D3D9 version
      • Allow hardware yuv->rgb conversions in setting dialog. Its' output:
      direct3d9 debug: Direct3d9 Device: AMD Radeon RX 580 2048SP 4098 28639 239
direct3d9 warning: trying surface pixel format: NV12
direct3d9 debug: selected surface pixel format is NV12
direct3d9 debug: Direct3D9 scene created successfully
direct3d9 warning: Direct3D shaders initialization failed !
direct3d9 debug: Direct3D9 device adapter successfully initialized
main debug: using vout display module "direct3d9"
main debug: A filter to adapt decoder DX11 to display NV12 is needed
main debug: looking for video converter module matching "any": 24 candidates
main debug: using video converter module "d3d11_filters"
main debug: Filter 'd3d11_filters' (0000023732721b50) appended to chain
main debug: original format sz 592x400, of (0,0), vsz 586x386, 4cc DX11, sar 1:1, msk r0x0 g0x0 b0x0

      I don't know how direct3d9 convert nv12 to display format, what type of color matrix and range type are currently using!

      • Disallow hardware yuv->rgb conversions
      direct3d9 debug: Direct3D9 scene created successfully
direct3d9 warning: Direct3D shaders initialization failed !
direct3d9 debug: Direct3D9 device adapter successfully initialized
main debug: using vout display module "direct3d9"
main debug: A filter to adapt decoder DX11 to display RV32 is needed
main debug: looking for video converter module matching "any": 24 candidates
chain debug: Trying to use chroma I420 as middle man
main debug: looking for video converter module matching "any": 24 candidates
main debug: using video converter module "d3d11_filters"
main debug: Filter 'd3d11_filters' (00000292c056a860) appended to chain
main debug: looking for video converter module matching "any": 24 candidates
swscale debug: 586x386 (592x400) chroma: I420 -> 586x386 (592x400) chroma: RV32 with scaling using Bicubic (good quality)
main debug: using video converter module "swscale"
main debug: Filter 'Swscale' (00000292c0568970) appended to chain
main debug: using video converter module "chain"
main debug: Filter 'chain' (00000292c0569730) appended to chain
main debug: original format sz 592x400, of (0,0), vsz 586x386, 4cc DX11, sar 1:1, msk r0x0 g0x0 b0x0

      It is complicated! ID3D11VideoContext1::VideoProcessorSetOutputColorSpace1, src_range and dst_range options of swscale module could help?

      1. OpenGL version

      I found that vlc is using matrix like matrix_bt601_tv2full and ignore range type information.

    • Jean-Baptiste Kempf
      Jean-Baptiste Kempf @jbk ·
      Owner

      You should disable hardware decoding, before testing.

      • Yafei Liu
        Yafei Liu @LiuYafei ·

        Ok, I'll try it again with D3D9. And the Video is not incorrectly displayed.

        modules log:

        main debug: Stream buffering done (500 ms in 0 ms)
avcodec debug: available hardware decoder output format 119 (cuda)
avcodec debug: available hardware decoder output format 53 (dxva2_vld)
avcodec debug: available hardware decoder output format 118 (d3d11va_vld)
avcodec debug: available hardware decoder output format 174 (d3d11)
avcodec debug: available software decoder output format 12 (yuvj420p)
main debug: trying to reuse free vout
main debug: removing module "freetype"
main debug: looking for text renderer module matching "any": 3 candidates
freetype debug: Using DWrite backend
freetype debug: DWrite_GetFamily(): family name: Arial
main debug: using text renderer module "freetype"
main debug: removing module "direct3d9"
direct3d9 debug: Direct3D9 scene released successfully
qt debug: IM: Deleting the input
main debug: Destroying the input for 'sc.mp4'
direct3d9 error: SetThumbNailClip failed: 0x800706f4
direct3d9 debug: Win32 Vout EventThread terminating
direct3d9 debug: Win32VoutCloseWindow
direct3d9 debug: WinProc WM_DESTROY
qt debug: IM: Setting an input
main debug: deinterlace -1, mode auto, is_needed 0
main debug: Opening vout display wrapper
main debug: looking for vout display module matching "direct3d9": 12 candidates
qt debug: IM: Deleting the input
qt debug: IM: Setting an input
direct3d9 debug: Using Direct3D9 Extended API!
direct3d9 warning: cannot load Direct3D9 Shader Library; HLSL pixel shading will be disabled.
direct3d9 debug: creating Vout EventThread
direct3d9 debug: Win32VoutCreateWindow
main debug: VoutDisplayEvent 'resize' 586x386
direct3d9 debug: inclinometer not found. (hr=0x80070490)
direct3d9 debug: created video sub-window
direct3d9 debug: Vout EventThread running
direct3d9 debug: disabling screen saver
direct3d9 debug: Direct3d9 Device: AMD Radeon RX 580 2048SP 4098 28639 239
direct3d9 warning: trying surface pixel format: DXA9
direct3d9 debug: selected surface pixel format is DXA9
main debug: VoutDisplayEvent 'resize' 602x425
direct3d9 debug: DirectXUpdateRects source offset: 0,0 visible: 586x386
direct3d9 debug: DirectXUpdateRects image_src coords: 0,0,592,400
direct3d9 debug: DirectXUpdateRects image_src_clipped coords: 0,0,586,386
direct3d9 debug: DirectXUpdateRects image_dst coords: 0,0,602,396
direct3d9 debug: DirectXUpdateRects image_dst_clipped coords: 0,0,602,396
direct3d9 debug: Direct3D created texture: 592x400
direct3d9 debug: Direct3D9 scene created successfully
direct3d9 warning: Direct3D shaders initialization failed !
direct3d9 debug: Direct3D9 device adapter successfully initialized
main debug: using vout display module "direct3d9"
main debug: original format sz 592x400, of (0,0), vsz 586x386, 4cc DXA9, sar 1:1, msk r0x0 g0x0 b0x0
main debug: reusing provided vout
main debug: looking for hw decoder module matching "none": 2 candidates
main debug: no hw decoder modules matched
main debug: VoutDisplayEvent 'resize' 586x386
direct3d9 debug: DirectXUpdateRects source offset: 0,0 visible: 586x386
direct3d9 debug: DirectXUpdateRects image_src coords: 0,0,592,400
direct3d9 debug: DirectXUpdateRects image_src_clipped coords: 0,0,586,386
direct3d9 debug: DirectXUpdateRects image_dst coords: 0,0,586,386
direct3d9 debug: DirectXUpdateRects image_dst_clipped coords: 0,0,586,386
main debug: removing module "direct3d9"
direct3d9 debug: Direct3D9 scene released successfully
direct3d9 error: SetThumbNailClip failed: 0x800706f4
qt debug: Logical video size: 586x386
main debug: resized to 586x386
main debug: VoutDisplayEvent 'resize' 586x386
direct3d9 debug: Win32 Vout EventThread terminating
direct3d9 debug: Win32VoutCloseWindow
direct3d9 debug: WinProc WM_DESTROY
main debug: deinterlace -1, mode auto, is_needed 0
main debug: Opening vout display wrapper
main debug: looking for vout display module matching "direct3d9": 12 candidates
direct3d9 debug: Using Direct3D9 Extended API!
direct3d9 warning: cannot load Direct3D9 Shader Library; HLSL pixel shading will be disabled.
direct3d9 debug: creating Vout EventThread
direct3d9 debug: Win32VoutCreateWindow
main debug: VoutDisplayEvent 'resize' 586x386
direct3d9 debug: inclinometer not found. (hr=0x80070490)
qt debug: Logical video size: 586x386
main debug: resized to 586x386
main debug: VoutDisplayEvent 'resize' 586x386
direct3d9 debug: created video sub-window
direct3d9 debug: Vout EventThread running
direct3d9 debug: disabling screen saver
direct3d9 debug: Direct3d9 Device: AMD Radeon RX 580 2048SP 4098 28639 239
direct3d9 warning: trying surface pixel format: NV12
direct3d9 debug: selected surface pixel format is NV12
main debug: VoutDisplayEvent 'resize' 602x425
direct3d9 debug: DirectXUpdateRects source offset: 0,0 visible: 586x386
direct3d9 debug: DirectXUpdateRects image_src coords: 0,0,592,400
direct3d9 debug: DirectXUpdateRects image_src_clipped coords: 0,0,586,386
direct3d9 debug: DirectXUpdateRects image_dst coords: 0,0,602,396
direct3d9 debug: DirectXUpdateRects image_dst_clipped coords: 0,0,602,396
direct3d9 debug: Direct3D created texture: 592x400
direct3d9 debug: Direct3D9 scene created successfully
direct3d9 warning: Direct3D shaders initialization failed !
direct3d9 debug: Direct3D9 device adapter successfully initialized
main debug: using vout display module "direct3d9"
main debug: A filter to adapt decoder DX11 to display NV12 is needed
main debug: looking for video converter module matching "any": 24 candidates
main debug: using video converter module "d3d11_filters"
main debug: Filter 'd3d11_filters' (000002729511d710) appended to chain
main debug: original format sz 592x400, of (0,0), vsz 586x386, 4cc DX11, sar 1:1, msk r0x0 g0x0 b0x0
main debug: reusing provided vout
main debug: looking for hw decoder module matching "none": 2 candidates
main debug: no hw decoder modules matched
main debug: VoutDisplayEvent 'resize' 586x386
direct3d9 debug: DirectXUpdateRects source offset: 0,0 visible: 586x386
direct3d9 debug: DirectXUpdateRects image_src coords: 0,0,592,400
direct3d9 debug: DirectXUpdateRects image_src_clipped coords: 0,0,586,386
direct3d9 debug: DirectXUpdateRects image_dst coords: 0,0,586,386
direct3d9 debug: DirectXUpdateRects image_dst_clipped coords: 0,0,586,386
main debug: removing module "d3d11_filters"
main debug: Filter 000002729511d710 removed from chain
main debug: removing module "direct3d9"
direct3d9 debug: Direct3D9 scene released successfully
direct3d9 error: SetThumbNailClip failed: 0x800706f4
direct3d9 debug: Win32 Vout EventThread terminating
direct3d9 debug: Win32VoutCloseWindow
direct3d9 debug: WinProc WM_DESTROY
main debug: deinterlace -1, mode auto, is_needed 0
main debug: Opening vout display wrapper
main debug: looking for vout display module matching "direct3d9": 12 candidates
direct3d9 debug: Using Direct3D9 Extended API!
direct3d9 warning: cannot load Direct3D9 Shader Library; HLSL pixel shading will be disabled.
direct3d9 debug: creating Vout EventThread
direct3d9 debug: Win32VoutCreateWindow
main debug: VoutDisplayEvent 'resize' 586x386
direct3d9 debug: inclinometer not found. (hr=0x80070490)
qt debug: Logical video size: 586x386
main debug: resized to 586x386
main debug: VoutDisplayEvent 'resize' 586x386
direct3d9 debug: created video sub-window
direct3d9 debug: Vout EventThread running
direct3d9 debug: disabling screen saver
direct3d9 debug: Direct3d9 Device: AMD Radeon RX 580 2048SP 4098 28639 239
direct3d9 warning: trying surface pixel format: YV12
direct3d9 debug: selected surface pixel format is YV12
main debug: VoutDisplayEvent 'resize' 602x425
direct3d9 debug: DirectXUpdateRects source offset: 0,0 visible: 586x386
direct3d9 debug: DirectXUpdateRects image_src coords: 0,0,592,418
direct3d9 debug: DirectXUpdateRects image_src_clipped coords: 0,0,586,386
direct3d9 debug: DirectXUpdateRects image_dst coords: 0,0,602,396
direct3d9 debug: DirectXUpdateRects image_dst_clipped coords: 0,0,602,396
direct3d9 debug: Direct3D created texture: 592x418
direct3d9 debug: Direct3D9 scene created successfully
direct3d9 warning: Direct3D shaders initialization failed !
direct3d9 debug: Direct3D9 device adapter successfully initialized
main debug: using vout display module "direct3d9"
main debug: original format sz 592x418, of (0,0), vsz 586x386, 4cc J420, sar 1:1, msk r0x0 g0x0 b0x0
main debug: reusing provided vout
avcodec warning: plane 0 not aligned: disabling direct rendering
main debug: VoutDisplayEvent 'resize' 586x386
direct3d9 debug: DirectXUpdateRects source offset: 0,0 visible: 586x386
direct3d9 debug: DirectXUpdateRects image_src coords: 0,0,592,418
direct3d9 debug: DirectXUpdateRects image_src_clipped coords: 0,0,586,386
direct3d9 debug: DirectXUpdateRects image_dst coords: 0,0,586,386
direct3d9 debug: DirectXUpdateRects image_dst_clipped coords: 0,0,586,386
main debug: Received first picture
main debug: Decoder wait done in 363 ms

        Should I disable ffmpeg hardware decoder too?

      • Jean-Baptiste Kempf
        Jean-Baptiste Kempf @jbk ·
        Owner

        Yes, you should

      • Yafei Liu
        Yafei Liu @LiuYafei ·

        I disabled hardware decoding in setting dialog, not sure if it is enough

      • Yafei Liu
        Yafei Liu @LiuYafei ·

        Or should I change can_hwaccel = true; to false ? It's not an easy task for me !

            for (size_t i = 0; pi_fmt[i] != AV_PIX_FMT_NONE; i++)
    {
        const AVPixFmtDescriptor *dsc = av_pix_fmt_desc_get(pi_fmt[i]);
        if (dsc == NULL)
            continue;
        bool hwaccel = (dsc->flags & AV_PIX_FMT_FLAG_HWACCEL) != 0;

        msg_Dbg( p_dec, "available %sware decoder output format %d (%s)",
                 hwaccel ? "hard" : "soft", pi_fmt[i], dsc->name );
        if (hwaccel)
            can_hwaccel = true;
    }
      • Jean-Baptiste Kempf
        Jean-Baptiste Kempf @jbk ·
        Owner

        No, preferences should be enough.

      • Yafei Liu
        Yafei Liu @LiuYafei ·

        Then last comment is exactly the result when hardware decoding is disabled.

      • Jean-Baptiste Kempf
        Jean-Baptiste Kempf @jbk ·
        Owner

        Weird, because you shouldn't have direct3d9 warning: trying surface pixel format: DXA9, but maybe I'm wrong.

      • Yafei Liu
        Yafei Liu @LiuYafei ·

        I noticed that too. And changing can_hwaccel = true; to false can give different log output

        avcodec debug: available hardware decoder output format 119 (cuda)
avcodec debug: available hardware decoder output format 53 (dxva2_vld)
avcodec debug: available hardware decoder output format 118 (d3d11va_vld)
avcodec debug: available hardware decoder output format 174 (d3d11)
avcodec debug: available software decoder output format 12 (yuvj420p)
main debug: looking for text renderer module matching "any": 3 candidates
freetype debug: Using DWrite backend
freetype debug: DWrite_GetFamily(): family name: Arial
main debug: using text renderer module "freetype"
main debug: looking for video converter module matching "any": 24 candidates
swscale debug: 32x32 (32x32) chroma: YUVA -> 16x16 (16x16) chroma: RGBA with scaling using Bicubic (good quality)
main debug: using video converter module "swscale"
main debug: looking for video converter module matching "any": 24 candidates
yuvp debug: YUVP to YUVA converter
main debug: using video converter module "yuvp"
main debug: Deinterlacing available
main debug: deinterlace -1, mode auto, is_needed 0
main debug: looking for vout window module matching "qt,any": 3 candidates
qt debug: requesting video window...
qt debug: IM: Setting an input
main debug: resized to 600x355
main debug: resized to 586x386
main debug: using vout window module "qt"
main debug: Opening vout display wrapper
main debug: looking for vout display module matching "direct3d9": 12 candidates
direct3d9 debug: Using Direct3D9 Extended API!
direct3d9 warning: cannot load Direct3D9 Shader Library; HLSL pixel shading will be disabled.
direct3d9 debug: creating Vout EventThread
direct3d9 debug: Win32VoutCreateWindow
main debug: VoutDisplayEvent 'resize' 586x386
direct3d9 debug: inclinometer not found. (hr=0x80070490)
direct3d9 debug: created video sub-window
direct3d9 debug: Vout EventThread running
direct3d9 debug: disabling screen saver
direct3d9 debug: Direct3d9 Device: AMD Radeon RX 580 2048SP 4098 28639 239
qt debug: Logical video size: 586x386
main debug: resized to 586x386
main debug: VoutDisplayEvent 'resize' 586x386
direct3d9 warning: trying surface pixel format: YV12
direct3d9 debug: selected surface pixel format is YV12
main debug: VoutDisplayEvent 'resize' 602x425
direct3d9 debug: DirectXUpdateRects source offset: 0,0 visible: 586x386
direct3d9 debug: DirectXUpdateRects image_src coords: 0,0,592,418
direct3d9 debug: DirectXUpdateRects image_src_clipped coords: 0,0,586,386
direct3d9 debug: DirectXUpdateRects image_dst coords: 0,0,602,396
direct3d9 debug: DirectXUpdateRects image_dst_clipped coords: 0,0,602,396
direct3d9 debug: Direct3D created texture: 592x418
direct3d9 debug: Direct3D9 scene created successfully
direct3d9 warning: Direct3D shaders initialization failed !
direct3d9 debug: Direct3D9 device adapter successfully initialized
main debug: using vout display module "direct3d9"
main debug: original format sz 592x418, of (0,0), vsz 586x386, 4cc J420, sar 1:1, msk r0x0 g0x0 b0x0
main debug: removing module "freetype"
main debug: VoutDisplayEvent 'resize' 586x386

        The result is still wrong I attach its' output too

      • Please register or sign in to reply
    • fafling
      fafling @fafling ·

      VLC version 3.0.17.4 Windows 10 64bit (up to date, DirectX End-User Runtimes (June 2010) is installed) Nvidia GeForce GTX 750 Ti (up to date driver with default settings, all video settings are "Use the video player setting")

      I have 3 possible workarounds for this issue in VLC "Tools > Preferences" menu :

      • Set "Video > Output" to "Direct3D 9 video output" (works with any value in "Input / Codec > Hardware accelerated decoding").
      • Set "Video > Output" to "DirectX (DirectDraw) video output" AND set "Input / Codec > Hardware accelerated decoding" to Disable or "DirectX Video Acceleration (DXVA 2.0)".
      • Set "Video > Output" to "OpenGL video output for Windows" or "OpenGL video output " AND set "Input / Codec > Hardware accelerated decoding" to "DirectX Video Acceleration (DXVA 2.0)".

      All other combinations of these options result in incorrect contrast for a full range video (when the video output works of course).

      Edited by fafling
    • Steve Lhomme
      Steve Lhomme @robUx4 ·
      Developer

      Can anyone provide logs with the default values ? "Video Output" and "Hardware decoder" to "Auto". This is what people should use.

    • Steve Lhomme
      Steve Lhomme @robUx4 ·
      Developer

      BTW Full Range YUV sources are incorrect. The BT 601/709/2020 define conversion matrices for limited range to full RGB. So if you use values out of the limited range (0-15, 235-255) in YUV they will be blown out in RGB. The D3D11 shader is trying to reduce the range (before clipping) but that's just a hack and not a mathematically correct operation.

      • Yafei Liu
        Yafei Liu @LiuYafei ·

        @robUx4 Correct me if I misunderstand what you mean. As far as I know, Full Range YUV sources are valid and correct in h.264 bitstream, video_full_range_flag syntax element is used to signal it. And I don't think the decoder can help us in this situation, it just decodes what bitstream says and doesn't care if the decoded value is in full/limited range.

        I am using vlc 4.0(SHA-1: 1f2e534f). Notice that we already have a variable named src_full_range in CompilePixelShader in direct3d11.c, since we are using matrices for limited range to full RGB anyway, then an extra step can be added to convert full range to limited range before colorspace conversion. Or using full range yuv to full range rgb matrix, there is a great article about how to calculate these matrices.

        Edited by Yafei Liu
      • Rémi Denis-Courmont
        Rémi Denis-Courmont @Courmisch ·
        Maintainer

        The decoder needs to provide the colour metadata to the filters and display, otherwise how would they know how to interpret the decoded pictures?

      • Yafei Liu
        Yafei Liu @LiuYafei ·

        Yes, I totally agree with you! I think VLC already can perfectly pass these information to d3d render. So I don't know why we are talking about decoders here? Or are you saying that decoders provide false info to later filters sometimes? That should be anther problem. Let's put it aside and assume that decoder does a good job, is it okay to use limited range to full RGB matrix even yuv is in full range? My knowledge is based on vlc 4.0 and d3d11, things could change if we are talking about different codes.

        Edited by Yafei Liu
      • Steve Lhomme
        Steve Lhomme @robUx4 ·
        Developer

        The official conversions I know are the ones defined in BT-601, BT-709 and BT-2020. They all define conversions from YUV limited range to RGB full range. They do not define how to convert full YUV to full RGB, nor limited YUV to limited RGB.

        The math used in the D3D11 shader come from this https://en.wikipedia.org/wiki/YCbCr#CbCr You can also notice they only have limited YUV to full RGB as well.

        It seems the avisynth conversions fall on the same values for limited YUV to full RGB (yuv [16,235] <-> rgb [0,255]). I'll see if their math can be used for full YUV to full RGB.

      • Please register or sign in to reply
    • Steve Lhomme mentioned in merge request !2327 (merged)

      mentioned in merge request !2327 (merged)

    • Jean-Baptiste Kempf closed with merge request !2327 (merged)

      closed with merge request !2327 (merged)

    • Jean-Baptiste Kempf changed milestone to %4.0

      changed milestone to %4.0

    • Jean-Baptiste Kempf added Status::fixed label

      added Status::fixed label

    Please register or sign in to reply

    VideoLAN code repository instance