Digital videos are everywhere in our day-to-day life from digital television, camera phones, Internet videos, video teleconferencing, car backup cameras, digital cinema, etc. Digital videos typically pass through many steps before they make it to your eyes & ears.

One important consideration is the displayed size. For example, there is no advantage in showing a 1920x1080 video on a mobile device with a 640x480 display. The video can be transformed to fit the desired display; thus, saving valuable bandwidth. Generally, this transformation is known as downsampling, and it consists of 3 parts:

• changing the resolution
• fixing the aspect ratio
• changing the bit-rate/sampling rate (for audio)

How do you know if the video quality is good after the transformation? The only reliable way to assess perceived video quality is to ask a group of people. However, this is really impractical due to time and cost. So companies like Video Clarity create Perceptual Video Quality algorithms to mimic the results of a human observer.

ClearView is our full-reference perceptual video quality analyzer. In order to predict the video quality, it compares the processed video to the original (reference/source). If the original and the processed are not the same size/shape, then the comparison is not valid.

So how do we measure the perceived video quality? Or more specifically, are we trying to check the quality of the downsampling? Or are we compressing after downsampling and want to know the video quality after compression?

If we are trying to compare the downsampling algorithm, then we could compare against common downsampling algorithms including bi-linear, bi-cubic, and Lanczos. Any of these can be performed using Virtual Dub. ClearView will predict the perceived quality using either the JND (Just Noticeable Differences) or DMOS (Differential Mean Opinion Score) scales. In either case, it does not know which video is better, it states the differences and then generates a score based on the size and scope of the differences.

If we are trying to compare video quality after downsampling and compression. This is quite easy. You need to perform the downsampling first and save this as the new source (reference). Then perform DMOS or JND measurements after compression.

We will be showing the effects of downsampling and upsampling during our upcoming trade shows. If you are in the area, please stop by for a discussion:

• VSF; Feb 2-4; Atlanta   -  free registration pass
• BVE; Feb 16-18; London
  
• CCBN; Mar 23-25; Beijing