# Wednesday, 24 March 2010

As I posted before I was looking for a bicubic resize algorithm for images to be used in Silverlight. I thought I ported the code properly from C# to VB but then yesterday when I was testing the code some more, I realized that there was a weird pattern of noise present in the picture. It looked like a square grid of sorts being overlaid onto the image. My experience told me that, that's because something screwed up when interpolating the pixel values (duh!) I didn't notice it previously because my previous test images didn't cause it to show up.

So what was I left to do? I had to port over the C# version to a lib on it's own then.

Bicubic Interpolation
/// <summary>
/// Bicubic resize algorithm suitable for use in Silverlight
/// </summary>
/// <param name="src"></param>
/// <param name="dest"></param>
/// <remarks>Original Source For This Function - http://www.codeproject.com/KB/recipes/aforge.aspx</remarks>
public static void resizeBicubic(System.Windows.Media.Imaging.WriteableBitmap src, WriteableBitmap dest)
{
    int height = src.PixelHeight;
    int width = src.PixelWidth;

    int newHeight = dest.PixelHeight;
    int newWidth = dest.PixelWidth;

    double xFactor = (double)width / newWidth;
    double yFactor = (double)height / newHeight;

    // coordinates of source points and cooefficiens
    double ox, oy, dx, dy, k1, k2;
    int ox1, oy1, ox2, oy2;
    // destination pixel values
    double r, g, b;
    // width and height decreased by 1
    int ymax = height - 1;
    int xmax = width - 1;

    int p;
    // RGB
    for (int y = 0; y < newHeight; y++)
    {
        // Y coordinates
        oy = (double)y * yFactor - 0.5f;
        oy1 = (int)oy;
        dy = oy - (double)oy1;

        for (int x = 0; x < newWidth; x++)
        {
            // X coordinates
            ox = (double)x * xFactor - 0.5f;
            ox1 = (int)ox;
            dx = ox - (double)ox1;

            // initial pixel value
            r = g = b = 0;

            for (int n = -1; n < 3; n++)
            {
                // get Y cooefficient
                k1 = BiCubicKernel(dy - (double)n);

                oy2 = oy1 + n;
                if (oy2 < 0)
                    oy2 = 0;
                if (oy2 > ymax)
                    oy2 = ymax;

                for (int m = -1; m < 3; m++)
                {
                    // get X cooefficient
                    k2 = k1 * BiCubicKernel((double)m - dx);

                    ox2 = ox1 + m;
                    if (ox2 < 0)
                        ox2 = 0;
                    if (ox2 > xmax)
                        ox2 = xmax;

                    // get pixel of original image
                   // p = src + oy2 * srcStride + ox2 * 3;
                    int srcPixel=src.Pixels[ox2+(width*oy2)];

                    r += k2 * ((srcPixel>>16) & 0xff);
                    g += k2 * ((srcPixel >>8) & 0xff);
                    b += k2 * (srcPixel & 0xff);
                }
            }

            
            dest.Pixels[x+(newWidth*y)]=(int)(0xff000000 | ((byte)r<<16) | ((byte)g<<8) |(byte) b);

        }
       
    }
}

public static double BiCubicKernel(double x)
{
    if (x > 2.0)
        return 0.0;

    double a, b, c, d;
    double xm1 = x - 1.0;
    double xp1 = x + 1.0;
    double xp2 = x + 2.0;

    a = (xp2 <= 0.0) ? 0.0 : xp2 * xp2 * xp2;
    b = (xp1 <= 0.0) ? 0.0 : xp1 * xp1 * xp1;
    c = (x <= 0.0) ? 0.0 : x * x * x;
    d = (xm1 <= 0.0) ? 0.0 : xm1 * xm1 * xm1;

    return (0.16666666666666666667 * (a - (4.0 * b) + (6.0 * c) - (4.0 * d)));
}

Pretty much a direct port of the original code since it was in C# already previously, just changed the way pixels are read and set. Same caveat applies as the previous code, no alpha premultiplication done yet.

The C# code didn't produce the weird grid like overlay pattern. So... what happened? Well my guess would have to be that I haven't properly ported the code to VB, and somewhere there's a fractional number being cast to a whole number causing some problems.

I'll have to be more careful when porting code that involves fractions in the future then.


Wednesday, 24 March 2010 10:08:14 (Malay Peninsula Standard Time, UTC+08:00)  #    Comments [0]  | 
# Monday, 22 March 2010

Additional 32GB storage, an extended battery that's HALF THE SIZE OF THE CAMERA, and a Gorillapod to be used in a manner which I don't really want to.

DSC06389

Yup... I'm all ready for the big event.

And then there's this.

DSC06390

Who would even guess that there's a hacked up mobile internet router in this thing?


Monday, 22 March 2010 00:31:05 (Malay Peninsula Standard Time, UTC+08:00)  #    Comments [0]  | 
# Sunday, 21 March 2010

The Sony CX-150 has a few nifty tricks up it's sleeve. First of all if you buy a cable from Sony, you can directly plug a HARD DISK into the camera and back up the internal memory's contents to said hard disk. Great for transfers during a vacation. Not so great since I tried using my own Mini-B to A Female cable and it didn't work, means the connection is proprietary and I don't want to pay a few hundred bucks for a simple cable.

The other trick the CX-150 can do is while you're recording a video you can just push this little button here.

DSC06365

The photo button, and it'll capture a photo while you're recording video. I'd have love to say you can capture a USABLE photo but I can't... heck I'm not even gonna justify it's existence by putting a sample photo here. It's VERY VERY BAD. It's great for people who aren't sure how to capture a frame off recorded video but other than that.. nope.. not usable at all.

The one last trick the CX-150 can do is pretty nifty... it can take high speed shots for roughly 5 seconds. It basically tries to be one of those high speed cameras. Problem is... you have to switch to this mode, and it can only take 5 seconds, and after taking the 5 seconds, it needs some processing time to make the video. Not so good for spur of the moment clips.

Great if you can setup a scene properly!

After taking the above video, I realized that the Vulcan couldn't fire fast enough (I was happy with the effect of course!) So I gave my brother the Raider to shoot with, but he wasn't experience enough to fire it fast enough. So I offered to shoot him instead, but he declined. I stared at my wife holding the camera but she scowled at me. Where would I find someone who takes delight in being shot by rapid firing Nerf darts? Oh... right... (Warning... scenes of offspring abuse follows)

I was aiming below the head! So I know I won't hit him in the face. :P


Sunday, 21 March 2010 21:51:00 (Malay Peninsula Standard Time, UTC+08:00)  #    Comments [1]  | 

I own 2 dedicated video cameras. My first true digital video camera was the Sony SR60.

DSC06376

I got it because at that time I knew as a father I would WANT to have something that can record proper video. And that the SR60 was pretty much the first camera I knew that was spitting out VBR MPEG2 files that I could just easily copy off the camera and manipulate. I definetly DIDN'T want to deal with tape, and I didn't want to deal with direct DVD cameras as well since I knew my videos final destination is the computer and NOT a DVD player.

My other video camera is the Creative Vado.

DSC06382

I grabbed it because I realized the need for a dedicated compact video device which I might want to carry around easily in my pockets. And I have used the Vado in EXTREME situations, I've taped it to a paintball gun before and I've also used it when I went for the SkyTrex experience, it's small size makes it easy for me to consider carry it wherever the need arises. While video is of usable quality, the mic pickup left a lot to be desired.

So I've bought tickets for a legendary musical event next month. And they WELCOME video recordings of their show, and I plan to make FULL USE of that invitation. But I don't know how thrill the venue people might be at the sight of a camera, so I thought about bringing the Vado, but considering that the Vado's audio pickup sucks and I couldn't afford stage side seats, I'd have to bring the SR60 then. But then I told myself, well... it's been awhile since I got the SR60, and there's this whole HD thing going on now. Why don't I indulge and take the opportunity to find a HD camera. I didn't want to spend too much on a pro grade video camera cause it's not like there's a reason I need CRYSTAL CLEAR video, so I was actually thinking of the Creative Vado HD 2nd Gen since from some sample videos it seemed to have better audio pickup.

Then I found this today...

DSC06360

This is the Sony CX-150, yes... I just bought YET ANOTHER Sony camera. This has EXTRA impact when I say it because nearly all (with the exception of the Creative Vado up there) of my Digital Cameras which I have used throughout my life HAS BEEN A SONY! Their cameras just have that extra something which I prefer I guess. So why did I pickup another full blown handy cam instead of my initial mini cam idea? Because the CX-150 is SMALL.

DSC06379

It's almost as long as the SR60 but the CX-150 is much thinner, the only reason it is as thick as it is seems to be to accommodate the lens assembly. Another reason why I choose this over a mini cam is that this is a FULL BLOWN video camera, means it comes with all the frills of a proper video camera ie. Decent audio pickup and OPTICAL ZOOM capability. So I don't sacrifice performance for size with the CX150. And for those of you complaining that it ain't THAT small. Well my definition of portable is that I can stick it into my Dockers Mobile Pant's pocket. ;)

The reason the CX150 can be so small is because it uses Sony Memory Stick Pro DUO (!) or SD HC (phew... class 4 and above only though) media cards for writing it's video to.

150door

But the camera doesn't come with any card because (as you could see in the first photo) it has an onboard flash memory of 16GB. Pretty decent, but it can only hold about 85 minutes on the camera's HIGHEST (it's some weird 1920x1080 24Mbps video, I say weird because it talks about video compatibility issues with hardware players when you switch to that setting) quality setting. Dropping the quality down one notch gives you 115 minutes, and dropping down another notch gives you 230 minutes. At the lowest setting the onboard memory can hold 6 hours worth of video.

But you wouldn't need to worry about that too much since the bundled battery...

DSC06366

Is good for about 125 minutes of continous recording time, LESS if you zoom in and cause it to auto focus during the recording of course. I guess I can force more usage time by turning off the screen during recording.

There's nothing much to talk about the design of the CX150, it's your typical fold out video camera design.

DSC06370

There's really nothing much to say about it. I do need to comment on one thing though. Even though the CX150 is very light, at no time does it feel CHEAP. In your hands it feels like a solid well made gadget. Wish Fujitsu paid attention to this detail when they were making the UH900.

One welcomed change from the SR60 is the fact that the CX150 has all important connection points on the camera itself. The power and analog video sockets on on the right.

DSC06367

Where as the STANDARD mini USB and HDMI connectors are on the left.

DSC06369

They didn't use any flimsy or removable rubber flaps to cover the ports, they made actual doors! I like this because my main gripe with the SR60 was that if I wanted to get files off the camera, not only did I needed to bring the docking station where the mini USB connector was. I also needed to bring the power adapter cause the SR60 won't power up without it! The CX150 has no problems acting as a mass storage drive when connected to the PC without the adapter. Though I don't think it actually draws any power from the USB port so if it's battery is dead you won't be able to connect it to your computer.

One welcome feature that DIDN'T change from my SR60 is the lens.

DSC06374

The CX150's lens design and heck.. I think the lens itself! Remains the same as the SR60. This means that the old wide angle lens I got for the SR60 fits like a dream on the CX150.

DSC06375

Obviously image quality suffers, but it's not like I care anyway I just want to be able to capture WZ when he's standing right in front of the camera! :P There is one little change to how the lens works though. The lens cover has to be MANUALLY opened and closed by a switch on the side

DSC06372

The SR60's lens cover was automatic, so I now need to remember to close the cover or risk lens damage!

I'll upload some video once I have more time to work on it (I managed to capture the Putrajaya Hotair Balloon Fireworks Display!) But I'd have to say the CX150 is a medium quality HD camera, so it can't hold a light to the video quality of higher end cameras. But I don't mind because I'm happy with the performance vs size that I'm getting.

Here's some advice to all of you planning to get a HD video camera.

  • HD files are HUGE! For example, at the highest quality setting 27 seconds of video resulted in a 75MB file.
  • Dealing with HD files are a NON TRIVIAL operation. Your computer might not have the right video codecs to decode the file, and even if you DID have the right video codecs, decoding H.264 requires a pretty beefy system, or a codec that performs very well.

I don't think I'll upload any FULL SIZED sample video (30 secs already = 75MB, I'd think I'd go over most 100MB limits). Just likely edit it and throw it up to YouTube.


Sunday, 21 March 2010 01:35:50 (Malay Peninsula Standard Time, UTC+08:00)  #    Comments [0]  | 
# Friday, 19 March 2010

UPDATE : This code doesn't work properly, go here for a properly working one.

I needed to resize an image in Silverlight and then save it out as a JPEG. I could have just used a transform on it then rendered it to a WriteableBitmap but the transforms were optimized for realtime performance and doesn't work too well when you need to maintain quality.

I stumbled upon a Bicubic Resize code in a library called AForge.Net and it was in C#, so it was simple enough for me to convert it to VB for me to use it in Silverlight. And since I DIDN'T find a lot of Bicubic Interpolation .Net resize code on the net, I'm gonna just put this copy up here.

As a bonus I kept on my comments in so you guys can see what was going through my head as I was working my way through the code.

Bicubic Interpolation
''' <summary>
  ''' Bicubically resize the src bitmap to the dest bitmap
  ''' </summary>
  ''' <param name="src"></param>
  ''' <param name="dest">Destination Bitmap, should be initialized to the requested size.</param>
  ''' <remarks>Original Source For This Function - http://www.codeproject.com/KB/recipes/aforge.aspx</remarks>    
  Sub Resize(ByVal src As Imaging.WriteableBitmap, ByVal dest As Imaging.WriteableBitmap)

      Dim startTime As DateTime = Now


      Dim srcwidth As Integer = src.PixelWidth
      Dim srcHeight As Integer = src.PixelHeight

      Dim destWidth As Integer = dest.PixelWidth
      Dim destHeight As Integer = dest.PixelHeight

      Dim xFactor As Double = srcwidth / destWidth
      Dim yFactor As Double = srcHeight / destHeight

      'Coordinates of src points and  coefficients.. WTF?
      Dim ox, oy, dx, dy, k1, k2 As Double
      Dim ox1, oy1, ox2, oy2 As Integer

      'new color values
      Dim r, g, b As Double

      Dim xmax As Integer = srcwidth - 1
      Dim ymax As Integer = srcHeight - 1

      For y As Integer = 0 To destHeight - 1
          'this is getting the original Y pixel if I'm not mistaken
          oy = y * yFactor - 0.5F
          'WHY is this necessary?
          oy1 = oy
          'WTF? It's getting the FRACTIONAL difference between
          'oy and oy1?
          dy = oy - oy1


          For x As Integer = 0 To destWidth - 1
              'same WTF as with the Y values,
              'maybe will make more sense later?
              ox = x * xFactor - 0.5F
              ox1 = ox
              dx = ox - ox1

              r = 0
              g = 0
              b = 0


              'this loop is to gather the interpolated
              'values of 2 pixels surrounding the current one
              For n As Integer = -1 To 2
                  'this gets the Y coefficient

                  k1 = BicubicKernel(dy - n)

                  'this seems to be getting the new
                  'Y pixel where the interpolated
                  'value comes from
                  oy2 = oy1 + n

                  'this is to ensure we're in the right spot
                  If oy2 < 0 Then
                      oy2 = 0
                  ElseIf oy2 > ymax Then
                      oy2 = ymax
                  End If

                  For m As Integer = -1 To 2
                      'for X coefficient
                      k2 = k1 * BicubicKernel(m - dx)

                      ox2 = ox1 + m

                      If ox2 < 0 Then
                          ox2 = 0
                      ElseIf ox2 > xmax Then
                          ox2 = xmax
                      End If

                      'get original pixel color
                      ' Dim origColor As Color = source.GetPixel(ox2, oy2)

                      'writablebitmap pixel is AARRGGBB
                      Dim srcColor As Integer = src.Pixels(ox2 + (srcwidth * oy2))


                      'set interpolated values
                      r += k2 * ((srcColor >> 16) And &HFF)
                      g += k2 * ((srcColor >> 8) And &HFF)
                      b += k2 * (srcColor And &HFF)
                  Next
              Next

              'after calculating coefficients we have our new color               
              dest.Pixels(x + (destWidth * y)) = &HFF000000 Or (r << 16) Or (g << 8) Or b


          Next
      Next
      dest.Invalidate()
      Dim endTime As DateTime = Now
      Diagnostics.Debug.WriteLine(String.Format("Time taken to squish {0}x{1} to {2}x{3} : {4} seconds", srcwidth, srcHeight, destWidth, destHeight, endTime.Subtract(startTime).TotalSeconds))

  End Sub

  Shared Function BicubicKernel(ByVal x As Double) As Double

      If x > 2.0 Then
          Return 0.0
      End If

      Dim a, b, c, d As Double
      Dim xm1 As Double = x - 1.0
      Dim xp1 As Double = x + 1.0
      Dim xp2 As Double = x + 2.0



      If xp2 <= 0 Then
          a = 0
      Else
          a = xp2 * xp2 * xp2
      End If

      If xp1 <= 0 Then
          b = 0
      Else
          b = xp1 * xp1 * xp1
      End If

      If x <= 0 Then
          c = 0
      Else
          c = x * x * x
      End If

      If xm1 <= 0 Then
          d = 0
      Else
          d = xm1 * xm1 * xm1
      End If
     
      Return (0.16666666666666666 * (a - (4.0 * b) + (6.0 * c) - (4.0 * d)))

  End Function

There's one caveat though, I didn't do any Alpha premultiplication with this code, which basically means if your images have a non FULLY opaque region, the results won't be pretty. Here's your homework assignment if you want to have that functionality.

I learnt 2 things while porting this code.

  1. Don't use Math.Pow() if your code is time sensitive, ie. it needs to work REALLY REALLY fast. Like in a game render loop. Takes a bit longer to run math.pow(x,3) instead of x*x*x. Not saying you shouldn't use it, just don't use it when MILISECONDS matter.
  2. Visual Basic's IIF statement isn't good for performance as well since BOTH true and false statements seem to be evaluated and does execution time is wasted when execution time matters. Better to use the usual IF..ELSE..ENDIF construct.

After my little exercise in dabbling with this, THEN I find a WriteBitmapEx library which pretty much seems to be a very useful class for raw manipulation of images in Silverlight, check it out!


Friday, 19 March 2010 00:24:20 (Malay Peninsula Standard Time, UTC+08:00)  #    Comments [0]  |