We normally use the center of the quads as the location of the tracking points, which is optimal for most cases and will account for a skew in the image. However, In some cases, a slight skew might not be real and instead be caused by image compression errors and/or low resolution. For those cases, we'd be better off centering the point exactly in the middle of the black area. This change computes the location data for the naively centered points as it is little additional work and allows for multiple attempts at decoding harder images.
An example of this bug:
This change fixes 5 of the 47 currently failing tests, with no regressions.
We normally use the center of the quads as the location of the tracking points, which is optimal for most cases and will account for a skew in the image. However, In some cases, a slight skew might not be real and instead be caused by image compression errors and/or low resolution. For those cases, we'd be better off centering the point exactly in the middle of the black area. This change computes the location data for the naively centered points as it is little additional work and allows for multiple attempts at decoding harder images.
An example of this bug:![buggycenter](https://user-images.githubusercontent.com/124665/79974176-b0190780-844d-11ea-8dbd-ada0217ba6f1.png)
This change fixes 5 of the 47 currently failing tests, with no regressions.