Disparity_DP.py

import cv2
import random
import numpy as np
import matplotlib

matplotlib.use("TkAgg")
from matplotlib import pyplot as plt

left_img = cv2.imread('output/calibration_binocular/rectifiedleft04.jpg',
                      cv2.IMREAD_GRAYSCALE)  # read it as a grayscale image
right_img = cv2.imread('output/calibration_binocular/rectifiedright04.jpg', cv2.IMREAD_GRAYSCALE)

# Disparity Computation for Left Image

# OcclusionCost = 20 (You can adjust this, depending on how much threshold you want to give for noise)
OcclusionCost = 20

rows, cols = left_img.shape

d_map_l = np.zeros(left_img.shape)
d_map_r = np.zeros(right_img.shape)


def optimal_match():
    for x in range(rows):
        CostMatrix = np.zeros((cols + 1, cols + 1))
        DirectionMatrix = np.zeros((cols + 1, cols + 1))

        for i in range(1, cols + 1):
            CostMatrix[i, 0] = i * OcclusionCost
            CostMatrix[0, i] = i * OcclusionCost

        for r in range(1, cols + 1):
            for c in range(1, cols + 1):
                min1 = CostMatrix[r - 1, c - 1] + np.abs(int(left_img[x, r - 1]) - int(right_img[x, c - 1]))
                min2 = CostMatrix[r - 1, c] + OcclusionCost
                min3 = CostMatrix[r, c - 1] + OcclusionCost

                CostMatrix[r, c] = min([min1, min2, min3])
                cmin = CostMatrix[r, c]

                if cmin == min1:
                    DirectionMatrix[r, c] = 1
                elif cmin == min2:
                    DirectionMatrix[r, c] = 2
                elif cmin == min3:
                    DirectionMatrix[r, c] = 3

        p = cols
        q = cols

        while p != 0 and q != 0:
            if DirectionMatrix[p, q] == 1:
                p = p - 1
                q = q - 1
                d_map_l[x][p] = np.abs(p - q)
                d_map_r[x][q] = np.abs(p - q)
            elif DirectionMatrix[p][q] == 2:
                p = p - 1
                d_map_l[x][p] = np.abs(p - q)
            elif DirectionMatrix[p][q] == 3:
                q = q - 1
                d_map_r[x][q] = np.abs(p - q)


def show_image(title, image):
    max_val = image.max()
    # image = np.absolute(image)
    image = np.divide(image, max_val)
    # cv2.imshow(title, image)
    cv2.imwrite(title, image * 255)


def main():
    optimal_match()
    print("========================Left Disparity=========================")
    print(d_map_l)
    print("========================Right Disparity=========================")
    print(d_map_r)
    show_image('output/Disparity/DPright04.jpg', d_map_l)
    show_image('output/Disparity/DPleft04.jpg', d_map_r)


main()