Stream local mjpg video to html canvas

前端 未结 2 1163
执笔经年
执笔经年 2020-12-07 05:03

I am trying to write a live stream of a mjpg video to an html canvas.

The following: http://camelive.info/ has a list of public webcams with mjpeg videos but they se

2条回答
  •  难免孤独
    2020-12-07 05:32

    According to specs about the CanvasRenderingContext2D drawImage method,

    Specifically, when a CanvasImageSource object represents an animated image in an HTMLImageElement, the user agent must use the default image of the animation (the one that the format defines is to be used when animation is not supported or is disabled), or, if there is no such image, the first frame of the animation, when rendering the image for CanvasRenderingContext2D APIs.

    This applies to .gif, SMIL animated .svg and .mjpeg media. So once you fetched the data, only one frame should be drawn onto the canvas.

    Note that chrome has a bug and only respect it for .gif images, but they may fix it someday.

    One solution as you noticed yourself, is to fetch an other fresh frame, with the clear-cache hack ('your.url/?' + new Date().getTime();) but you will loose any advantages of the mjpeg format (partial frame content) and can't be sure when the refreshing will happen.

    So a better solution if applicable, would be to use a video format. Each frame of a video can be drawn to the canvas.


    Edit 2018


    A third solution came to my little mind two years later:

    UAs are not tied to keep in memory the same default image for all 2DContexts in the document.
    While for others format we are still kinda stuck, for MJPEG streams, which don't have a well defined default image, we actually fall to the first frame of the animation.

    So by drawing the containing our MJPEG stream on two different canvases, at different times, we can theoretically have two different frames of our same MJPEG stream to be drawn on the canvases.

    Here is a proof of concept only tested on Firefox 62.

    var ctx_stream = stream.getContext('2d');
    var ctx_direct = direct.getContext('2d');
    img.onload = function() {
       stream.width = direct.width = this.naturalWidth;
       stream.height = direct.height = this.naturalHeight;
       // onload should fire multiple times
       // but it seems it's not at every frames
       // so we'll disable t and use an interval instead
       this.onload = null;
       setInterval(draw, 500);
    };
    function draw() {
      // create a *new* 2DContext
      var ctx_off = stream.cloneNode().getContext('2d');
      ctx_off.drawImage(img, 0,0);
      // and draw it back to our visible one
      ctx_stream.drawImage(ctx_off.canvas, 0,0);
      
      // draw the img directly on 'direct'
      ctx_direct.drawImage(img, 0,0);
    }
      
      
    img.src = "http://webcam.st-malo.com/axis-cgi/mjpg/video.cgi?resolution=704x576&dummy=1491717369754";
    canvas,img{
      max-height: 75vh;
    }
    Using a new offcreen canvas every frame: 

    The original image:

    Drawing directly the <img> (if this works your browser doesn't follow the specs):

    So while this solution will obviously come with a performance impact (we are creating a whole new canvas element and its 2DContext every frame), it's still probably better than flooding the network. And all this should be Garbage Collected quite easily anyway.

提交回复
热议问题