I use
worldview_inverse * (projection_inverse * vector)
to convert the coordinates of the screen space to world space coordinates. I suggested that
(x,y,1,1)
is converted to a point in the background, and
(x,y,-1,1)
converts to a point on the near plane, and by connecting the line, I can query all the objects in a truncated representation that intersect the line. After the conversion, I divide the resulting result points into their corresponding .w component. This works for the far plane, but the point on the near plane is somehow transformed into world cosmic origin.
I think this is due to the components of w 1, which I feed into the back projection, because usually it is 1 before the projection, and not after, and I do the back projection. What am I doing wrong?