Sony S9000ES manual Block Noise Reduction

Sony’s Motion Adaptive Field Noise Reduction uses new technology to overcome the old compromises. Thanks to the new circuit, the DVP-S9000ES can achieve very high signal-to- noise ratio, very high resolution and very accurate motion

— all at the same time!

Fig. 6: Sony analyzes the video fields for movement by dividing the fields into blocks that are 4 pixels high by 16 pixels wide. When big changes are detected between two fields, Sony’s MPEG Image Processor determines that there is screen movement. The system attempts to match the block across a search window that extends four pixels up, three pixels down, eight pixels to the left and seven to the right.

To overcome motion artifacts, the Sony design actively analyzes the fields for movement. The system divides the image into blocks of 4 pixels high by 16 pixels wide. This specific rectangular shape is based on the understanding that most movement in video images is horizontal. In blocks where no large changes are detected between fields, the system applies conventional noise reduction.

When large changes are detected between fields, the system automatically searches for block movement. The search “window” is 11 pixels high by 31 pixels wide.

By finding matches for blocks that have moved, Sony’s Motion Adaptive Field Noise Reduction makes it possible to apply noise to areas of screen movement, in addition to the still background. The system compares data from the two fields by a mathematical method called the Hadamard transform. Noise, in the form of

Fig. 7: Sony’s Motion Adaptive Field noise reduction identifies and suppresses noise across the entire screen area — in moving parts of the picture as well as the motionless background. The process uses advanced motion detection and a Hadamard transform to suppress noise without sacrificing detail or generating motion artifacts.

differential luminance (Y) signals, is analyzed in blocks one pixel high by 8 pixels wide. The Hadamard transform converts the noise into an easily processed frequency distribution. To reduce errors, eight operations are performed for each pixel. The average of the eight values is then used. A limiter is then used to extract noise from the converted signals. Than a reverse Hadamard transform generates an error-correction signal that is added to the video signal to suppress noise.

In addition, Sony’s Motion Adaptive Field Noise Reduction performs the same process for noise in the color difference channels, CB and CR. Since luminance and color difference signals are separately processed, their noise reduction can be individually optimized.

The mathematics of this noise reduction process may be complex, but the results are easy to appreciate. The circuit accomplishes three formerly elusive goals simultaneously:

1.Clean noise-free images on both moving portions and still portions of the screen.

2.The full resolution of the DVD.

3.Clear and natural image movement, without ghosting or other motion artifacts.

Block Noise Reduction

Another characteristic limitation of MPEG-2 compression is block noise. This is the tendency for delicately shaded picture areas to be rendered as solid rectangles of color. Instead of a continuous contour of color on the cheek of an actress, fading softly into shadow, you see subtle squares of approximately correct color. Block noise is most apparent at the edges where squares meet. This effect is also called mosquito wings, because it can appear as subtle discoloration in tiny slices of the picture and just as quickly disappear.

Sony’s MPEG Image Processor mounts a sophisticated, compre- hensive attack on block noise. The aim of Sony’s system is to identify those areas of subtle gradation of tone — the areas most prone to block noise — analyze the gradations and reconstruct the abrupt steps of tone as more linear, more gradual slopes.

The principal challenge here is to distinguish legitimate steps in the picture tone (signal) from the unwanted, artificial steps (block noise). This separation of wanted and unwanted steps is made easier because the block borders area always at the same places on the screen. Their location is a fixed attribute of the MPEG-2 compression used in the DVD format.

The DVD format divides the screen into 2,700 blocks (90 hori- zontal by 30 vertical). Each block measures 8 pixels horizontal by 8 pixels vertical. They’re shown as vertical rectangles because of interlace scanning. To judge for block noise, the MPEG Image Processor establishes the three pixels to the left and right of the

8