With the recent launch of our next generation video cores, the PowerVR D4500MP decoder and PowerVR E4500MP encoder, we have brought a series of exciting new features designed to preserve colour fidelity from source to display and provide the performance required for ultra-high definition video applications. Building on the strength of the highly acclaimed PowerVR Series3 IP core families, these multi-pipe video IPs will usher in a new era of colour quality and fidelity by offering a hardware coding engine capable of supporting full chroma resolution up to 10 bits of precision.
So why is it so important to keep the same colour quality throughout our internal processing pipeline? Well, a few years ago the question might have been relevant only to graphics enthusiasts, digital artists and graphics professionals, whereas the rest of the planet would think find the classic “True colour” option to be suitable enough for their needs. But “suitable enough” was never good for us. And we are not alone.

Video transcoding with 10-bit precision – capturing the world in 1 billion colours

In answering this dilemma, we first need to think about the concept of representing an image as a matrix where each element is called a pixel. Each pixel has a set of three values, depending on which colourspace is used. There are a number of possibilities but two have been extensively used because they come very close to representing the way we perceive colour: either as a combination of red, green and blue (RGB) or as a mix of luminance (brightness) and chrominance (colour) (YCbCr). The larger the interval of possible values, the better the pixel’s digital value can approximate the real-life colour. “True colour” means one can have 8 bit values for each pixel element, leading to 256 shades of red, green and blue which translates to almost 16 million colour variations. This depth has been traditionally used on the assumption that the human eye can distinguish between 10 million colours but the main problem is that these images do not simply travel from one point to another in a computing system. They are exposed to various types of processing, like enhancements, corrections and level adjustments to eliminate certain errors or defects but also encoding or decoding to limit bandwidth and allow for standardization. In going through the motions of these repeated and frequent transformations, images often lose quality and suffer deformations which become very visible (and most of the times annoying) to the human eye.

Show me a moviegoer who hasn’t felt the frustration of observing the banding effects on sky scrapers before they get smashed into pieces in the latest installment of the Dark Knight, Avengers or Transformers franchise. Or the disappointment of coming back from the best holiday you’ve ever had and noticing the complete lack of life in your videos and photos even though you can still remember how wonderful these places really looked. Below is an example of how reducing colour quality in compression can affect images or videos:

PowerVR video: banding effects in 8bit images

Banding effects in video encoding, observable particularly around the skyline*

PowerVR Series4 video IP maintains and improves image quality

Let’s take a quick glance at what happens to a video clip when it is being compressed according to the one of the most commonly used standards currently available: H.264/MPEG-4 AVC. For the encoding process, each video frame (essentially, a picture) is turned into a compact stream of data that can be efficiently stored in local memory. When the video needs to be played back, that stream is used to assemble a series of video frames that are output on a display.

 

H.264 video encoding and decoding block diagram

 

A majority of the operations performed are based on complex mathematical calculations. This is where colour precision starts to make a difference as many algorithms will be able to achieve nearly lossless coding by taking advantage of the higher colour precision. It might seem hard to believe but those 2 extra bits make all the difference, especially in the quantization and transform stages where a lot of information is lost due to rounding.

Simply put, through encoding, portions of the colour range are chopped off. Just have a look at the picture below and see for yourselves:

 

Differences in banding for 8bit and 10bit color spaces

With 10-bit colours, you can have an amazing total of 1024 shades for each primary colour. When you compare that to just 256 possible values for 8-bit colours, you can see why the new PowerVR video cores will give you a sense of depth and reality that can’t be matched by anything on the market right now.

Adding increased colour subsampling

Furthermore, the new PowerVR video cores have increased the colour resolution to 4:4:4. This means that when dealing with images saved in an YCbCr format, there is no need to perform chroma subsampling as each component has the same sample rate (for every Y value, there is a corresponding Cb and a Cr value). Thus, the conversion back to the RGB space is done using a simple set of formulas according to the coding standard in use. In contrast, implementing just a 4:2:0 sampling rate can have unwanted results like motion artefacts, comb-like or blur and trailing effects. The main issue is that for sharp colour transitions, which are particularly noticeable with graphics, blurring will occur in a 4:4:4-to-4:2:0-to-4:4:4 conversion process due to the interpolation needed. This is made worse, if this conversion is repeated several times in the output chain. Even with advanced post-processing, the original quality is forever lost and images become either too grainy or misty even to the untrained eye.
And it’s not just the quality itself. There are a number of other practical reasons why you should take advantage of the 10 bit range supported by the PowerVR video IP family. The next time you walk into a movie theatre showing digital media or sit in front of your TV playing something off a Blu-ray disc, think what it would be like to be able to distinguish clearly between shadows and highlights, especially in dark or poorly-lit environments, where everything tends to get a bit fuzzy. Now you can finally see that Alien slither down corridors in its full-fledged glory without the annoying banding prevalent today!
Finally, with more and more 4K content starting to appear and digital TV sets getting higher resolutions and larger form factors, a higher colour depth will be able to cope with the ever-increasing display surface.
The PowerVR D4500MP and E4500MP cores were designed with the latest industry trends in mind, including advancements in OLED technology, the addition wireless capabilities for screens and the expanding range of video applications. They manage to provide the edge needed for professionally and user generated content to be enjoyed the way we would want it: in full detail, with true to life colours and without missing a single frame.

Stay tuned to our blog as we will discuss more about our PowerVR video and display technology in the near future. Also, follow our Twitter account for more news and updates on the entire PowerVR family as well as Imagination’s Meta processors, Ensigma Communications platforms, and HelloSoft and Caustic Professional products.

* Image courtesy of Fringe Focus, all rights reserved

AHM4GD5WXMQC

About the author: Alex Voica

Profile photo of Alex Voica

Before deciding to pursue his dream of working in technology marketing, Alexandru held various engineering roles at leading semiconductor companies in Europe. His background also includes research in computer graphics and VR at the School of Advanced Studies Sant'Anna in Pisa. You can follow him on Twitter @alexvoica.

View all posts by Alex Voica