Commit f7ab7e37 authored by Sarah Parker's avatar Sarah Parker
Browse files

Update global motion experiment comments

Some of these comments are inconsistent or outdated.

Change-Id: Icf14996a8792cc7f2029a22100d0994a852306b5
parent b7e72f38
......@@ -120,13 +120,17 @@ static INLINE void set_default_warp_params(WarpedMotionParams *wm) {
// ALPHA here refers to parameters a and b in rotzoom model:
// | a b|
// |-b a|
// The following constants describe the various precisions
// of different parameters in the global motion experiment.
// and a, b, c, d in affine model:
// | a b|
// | c d|
// Given the general homography:
// [x' (a b c [x
// z . y' = d e f * y
// 1] g h i) 1]
// Constants using the name ALPHA here are related to parameters
// a, b, d, e. Constants using the name TRANS are related
// to parameters c and f.
// Anything ending in PREC_BITS is the number of bits of precision
// to maintain when converting from double to integer.
......@@ -37,13 +37,15 @@ double refine_integerized_param(WarpedMotionParams *wm,
Computes "num_motions" candidate global motion parameters between two frames.
The array "params_by_motion" should be length 8 * "num_motions", where the
first 2 slots in each group of 8 parameters are the translation parameters in
(row, col) order, and the remaining slots correspond to values in the
transformation matrix of the corresponding motion model. They are arranged in
"params" such that values on the tx-matrix diagonal have odd numbered indices
so the folowing matrix: A | B C | D would produce params = [trans row, trans
col, B, A, C, D].
The array "params_by_motion" should be length 8 * "num_motions". The ordering
of each set of parameters is best described by the homography:
[x' (m2 m3 m0 [x
z . y' = m4 m5 m1 * y
1] m6 m7 1) 1]
where m{i} represents the ith value in any given set of parameters.
"num_inliers" should be length "num_motions", and will be populated with the
number of inlier feature points for each motion. Params for which the
num_inliers entry is 0 should be ignored by the caller.
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment