Video Quality Tested: GeForce Vs. Radeon In HQV 2.0

Test Class 4: Adaptive Processing

The adaptive processing tests reveal the video processor’s ability to optimize contrast and color correction.

Chapter 1: Contrast Enhancement Tests

This chapter is composed of four tests, comprised of four different video clips that present different challenges for contrast enhancement. The first is a theme park during the day, the second is an overcast beach scene with driftwood, the third is a tropical beach at dusk, and the final scene has black and white cats laying beside each other. A perfect score of five is awarded for each scene where the contrast is expanded without loss of detail in the darkest and lightest regions. The score is lowered to two if there is slight loss of detail in the lightest and darkest regions, and a score of zero is given if there is moderate or high loss of detail.

This is one of the easier tests to score because it isn’t hard to notice if contrast has improved without loss of detail. It turns out that dynamic contrast is a little resource-intensive, and the low-end graphics hardware can’t handle the feature without choppy playback. But every card that is able to enable the dynamic contrast option scores a perfect five points for each test. We do notice that we have to disable the brighter whites option on Radeons cards so that the lightest regions will not overexpose and lose detail.

Chapter 2: Skin Tone Correction Tests

Human beings tend to be sensitive to unrealistic skin tones, so this test examines the video processor’s ability to detect and correct skin tones that are unrealistic. The test consists of a picture of people with various skin colors, and the skin tone hues are shifted off of their true colors over time. The maximum 10 points are awarded if off-hue skin tones are corrected to appear substantially closer to the original skin tone without affecting other colors in the scene. This drops to seven points if the skin tones are corrected somewhat, but problems in hue are still discernable, or three points if the skin tones are somewhat corrected, but non-skin colors are affected. If no improvement is observed, no points are awarded:

The Radeons have the advantage here, with the only dedicated flesh-tone setting in the driver. Even then, the test is admittedly difficult to judge. We could not detect any change when turning the GeForce color enhancement feature on while assessing the hardware.

[EDIT: We want to mention here that we've received some comments about skin tone correction noting that it can potentially mess with the way the director wants colors to look. This is certainly a valid argument, and there are those of us who will prefer to leave this setting off. However, this remains a valid comparative test as some video processors are equipped with this feature]