K-WILL eliminates weaknesses of traditional subjective testing. Using a patented three-layered noise weighting technology, K-WILL provides the same results as the attentive human eyes. K-WILL’s VP Series enables automated, accurate and real time video and audio quality analysis for every type of source material on a frame-by-frame basis. The test result for each frame is calculated and plotted in a graph with different colors assigned to highlight different value levels.
3-Layered Noise-Weighting Evaluation
In order to provide users with DSCQS values, K-WILL employs a patented 3-layered noise-weighting algorithm that calculates video and audio quality just as human visual perception.
- Stage One
Based on frame activity, we infer the object complexity of a frame and determine a general noise sensitivity value. This is based on the concept that higher object complexity causes weaker focus of the viewer’s attention (gaze points) and lower sensitivity to video noise. - Stage Two
Based on block activity, we infer the degree of “noise masking” effect within specific regions (= 8 X 8 small blocks) and determine the degree of noise sensitivity of each section within the regions. Human eyes tend to be sensitive to video noise that is overlaid on flat or smooth patterns where block activity is low, but on the other hand, they are less sensitive to noise that occur on images such as forests and trees as the images “mask the noise”. - Stage Three
Noise is segregated into high and low frequency ranges by applying a band-pass-type visual filter.
High Correlation with Subjective Testing Results
Video quality testing by human observers tends to vary in the test results. Namely, the correlation of the results of “subjective” tests will not necessarily be 1.00. The diagram below shows a comparison of test results using K-WILL’s VP products and those based on two executions of “subjective” testing. The correlation among the three test results was almost identical, demonstrating VP’s ability to produce test results that are comparable to traditional “subjective” testing.
Automatic Synchronization of Source and Tested Videos
In order to compare the source and the tested videos, it is necessary to accurately synchronize the two video signals. While conventional method attempted synchronization by inserting a special timing marker in the picture, KWILL’s products achieve synchronization based on the video signal itself (“markerless synchronization”). As a result, it is possible to automatically synchronize and analyze video images even as they are transmitted through broadcasting networks.
The possible range of delay compensation for synchronization is within 60 frames with a horizontal value of ±24 pixels and vertical value of ±4 lines.
What is DSCQS?
Video quality has traditionally been measured by human observers. Known as DSCQS (Double Stimulus Continuous Quality Scale), this method has been standardized under ITU-R (International Telecommunications Union-Recommendation) BT-500. Under this method, picture quality is expressed as a percent from 0 to 100, with 0 indicating no distortion. The higher the value, the greater the distortion between the original and the tested video images.
While DSCQS provides a good index for quality, this method requires not only a special testing room, monitors, and video editing but also a panel of 15 or more human observers. Inevitably, this type of testing has various limitations such as the inability to reproduce consistent results and conduct “real-time” evaluation as well as the difficulty in conducting extended testing. With K-WILL’s technology, it becomes easy to automate this process.