I received a 1920x1080 video transfer from a 35mm cine print. The transfer involved a chain of steps which the film's producer-director no longer remembered. The video had a fuzzy black border asymmetrical around the image: about 26 pixels on the left; about 40 on the right; about 10 above; about 26 below. So the image area was almost exactly 16:9. But the image was distorted: everything appeared too wide. I measured some circular things and found their widths about 1.121× their heights. This means that the image area should correctly be displayed at about 1.58:1. The film was released as 1.66:1, so some of its width had been cropped. What's cropped is lost. I decided to undistort the image and display it as 1.58:1 pillarboxed in standard 16:9 video.
The geometry pane of Compressor allows this. The geometry pane has three entry sections named "Source Inset (Cropping)", "Dimensions (encoded pixels)", "Output Image Inset (Padding)" for settings. The relation of the three is simple, but lacks clear description in the verbose User Manual.
..."Cropping" refers to pixels removed from the input image.
..."encoded pixels" refers to the whole output image.
..."Padding" refers to black edges of the whole output image.
The relation of the three is this:
...The input image area minus its crop is mapped onto the output image area minus its padding.
So I entered 26, 40, 10, 26 (as described) for crop and 1920, 1080 1920x1080 for encoded pixels. Square pixels. To find the padding then required arithmetic. I put 105 on the left and 104 on the right.
Dennis Couzin
Berlin, Germany