If you can I'd use IPublishedContent rather than IMedia as it deals better with the abstractions of getting a URL to the media. Once you have the relative URL you can map that to an absolute path on disk - not sure how this works for Azure blob as I don't use it - but it works for stuff in media folder. Once you have an absolute path you can read it into a stream and copy it to a MemoryStream.

Something like:

var media = Umbraco.TypedContent(1234); // your media Id

if (media != null)
{
    string path = Server.MapPath(media.Url);

    if (System.IO.File.Exists(path))
    {
        using (MemoryStream memoryStream = new MemoryStream())
        {
            using (Stream input = System.IO.File.OpenRead(path))
            {
                input.CopyTo(memoryStream);
            }

            memoryStream.Position = 0;

           // you can access your memory stream here
        }

    }
}

Finally, I use these functions to get a stream:

    private IMedia GetMedia(int id)
    {
        var context = ApplicationContext.Current;
        var services = context.Services;
        var mediaService = services.MediaService;

        return mediaService.GetById(id);
    }

    private ToStream(IMedia media)
    {
        var ms = new MemoryStream();

            var umbracoFile = media.GetValue<string>(Constants.Conventions.Media.File);
            var url = JsonConvert.DeserializeObject<ImageCropDataSet>(umbracoFile).Src;

            var path = Server.MapPath(url);

            if (System.IO.File.Exists(path))
            {
                using (Stream input = System.IO.File.OpenRead(path))
                {
                    input.CopyTo(ms);
                }
            }


        ms.Position = 0;
        return ms;
    }

But, it cannot retrieve image stored as blob using plugin UmbracoFileSystemProviders.Azure.

is working on a reply...