Image Processing Computer Vision Projection Model Stereo Vision

Image Processing & Computer Vision Projection Model & Stereo Vision

Projection Model l Computer Graphic (X, Y, Z) = 3 D Coordinate (X, Y) = 2 D Coordinate

Perspective Projection Model ���������� dispa

Stereo Vision Left Right Disparity (Ground Truth)

Stereo Vision Left Right Disparity (Ground Truth)

Stereo Vision : Disparity l Finding disparity Left Right Disparity

Constraints Data Constraints ���� � ��������������������� (image intensity) ��������� 2. Smoothness Constraints disparity ������ smooth ����� disparity ���������� (neighbor) R L 2+ Data Constraints Energy = [(I – I ) xy (x, y)) (x+D(x, y) ������������� 2 1. (D(x+1, y) – D(x, y)) (D(x, y+1) – D(x, y))2 ] + + Smoothness Constraints

Algorithm using Gibbs Sampler 1. 2. 3. Start Temperature T is high Initialize D(x, y) = Random 0…. 20 For each pixel(x, y) For each state S = 0… 20 if D(x, y) = 0; E 0 = … ; P 0 = exp(-E 0/T) if D(x, y) = 1; E 1 = … ; P 1 = exp(-E 1/T) ……………. if D(x, y) = 20; E 20 = … ; P 20 = exp(-E 20/T) For each Probi = Pi / sum(Pi) 4. Sample for state S from pdf Probi D(x, y) = State S 5. 6. Reduce T = T * 0. 9 Repeat step 3 -4 Until E is stable

Example Random disparity left right Result disparity

Example right left 0 1 2 3 4 5 6 0 ������ D(x, y) 1 4 state ������ (1, 2) (3, 3)��� 2 (2, 5) 3 4 5 6 Random disparity Result disparity

Data Constraint Trick (IR(x-D(x, y) – IL(x, y))2 �������� pixel ������ 9 pixels 1 1 (IR(x-D(x, y)+m, y+n) – IL(x+m, y+n))2 m= -1 n= -1 3 x 3 pixel

Display disparity in Grayscale (Example( Disparity ��� 12. 75 Intensity