Disparity computation

Theoretical basics

The disparity computed by Pandora2D is such that:

$I_{L}(x, y) = I_{R}(x + dx, y + dy)$

with $I_{L}$ , $I_{R}$ the left image (left image) and the right image (right image), and $dx$ the column disparity and $dy$ the row disparity.

At this stage, a 4D (dims: row, col, disp_col, disp_row) cost_volumes is store. We use the Winner-Takes-All strategy to find the right disparity for each pixel. That’s mean we are looking for the min (resp: max for zncc measure). For column’s disparities (resp: row’s disparities) we search the min or max in disp_row (res: disp_col) to obtain a 3D cost_volume (row, col, disp_col (res: disp_row)). To conclude, we extract the disparity of min (or max) from the 3D cost_volume and we obtain two disparity maps for row and col.

Configuration and parameters

Name	Description	Type	Default value	Available value	Required
disparity _method	Disparity method	string		“wta”	Yes
invalid_disparity	Invalid disparity value	str, int, float	NaN	“NaN”, “inf”, int	No

Example

{
    "input" :
    {
        // input content
    },
    "pipeline" :
    {
        // ...
        "disparity":
        {
            "disparity _method": "wta",
            "invalid_disparity": "NaN"
        },
        // ...
    }
}