Introduction

This repository aims to compute the disparity map between the left and the right image with differences in brightness via some common method L1, L2, Cosine Similarity and Correlation Coefficient

Getting started

• There are total of 5 problems:
  • Using L1 and L2 with pixel-wise between 2 images to get the disparity map
  • Using L1 and L2 with RoI represented with kernel size value to get the disparity map
  • Apply to the left and the right image with different brightness to evaluate performance of L1 and L2
  • Using Cosine similarity with RoI represented with kernel size value to get the disparity map from the left and the right image with different brightness
  • Using Correlation coefficient instead of Cosine similarity and compare the achieved results
• In addition, with the problems using RoI, i apply an integral mechanism to optimize a computational cost. You can see what it is with the following image.

Results

Method	disparityRange	Kernel	Brightness	Result
L1	16	None	Same
L1	64	None	Same
L2	16	None	Same
L2	64	None	Same
L1	16	5	Same
L1	64	5	Same
L2	16	5	Same
L2	64	5	Same
L1	64	3	Light
L2	64	3	Light
Cosine	64	3	Light
Cosine	64	5	Light
Cosine	64	3	Dark
Coef	64	3	Light
Coef	64	5	Light
Coef	64	3	Dark