Panorama.py

import cv2
import numpy as np


def RANSAC(src_pts, dst_pts, num_iterations, threshold):
    best_homography = None
    best_inliers = []
    for i in range(num_iterations):
        # Randomly select 4 pairs of matched points
        random_indices = np.random.randint(0, len(src_pts), 8)
        random_src_pts = src_pts[random_indices]
        random_dst_pts = dst_pts[random_indices]

        # Estimate the homography matrix using the random points
        homography_matrix, _ = cv2.findHomography(random_src_pts, random_dst_pts)


        # Calculate the distances between the transformed source points and the destination points
        dst_pts_transformed = cv2.perspectiveTransform(src_pts, homography_matrix)
        difference = dst_pts - dst_pts_transformed

        distances = np.linalg.norm(difference, axis=1)

        # Select the inliers based on the threshold
        inliers = np.where(distances < threshold)[0]
        # print(inliers)
        # Update the best homography and inliers if the current set is better
        if len(inliers) > len(best_inliers):
            best_homography = homography_matrix
            best_inliers = inliers
        else:
            np.delete(src_pts, random_indices)
            np.delete(dst_pts, random_indices)

    return best_homography, best_inliers




# Load the two images
img1 = cv2.imread("2.png")
img2 = cv2.imread("1.png")

# Initialize SIFT detector
sift = cv2.xfeatures2d.SIFT_create()

# Compute the keypoints and descriptors for both images
kp1, des1 = sift.detectAndCompute(img1,None)
kp2, des2 = sift.detectAndCompute(img2,None)

# Use BFMatcher to find the best matches
bf = cv2.BFMatcher()
matches = bf.knnMatch(des1,des2, k=2)

# Filter the matches based on their distance
good = []
for m,n in matches:
    if m.distance < 0.9*n.distance:
        good.append(m)

# Draw the matches using drawMatches function
img3 = cv2.drawMatches(img1,kp1,img2,kp2,good,None,flags=cv2.DrawMatchesFlags_NOT_DRAW_SINGLE_POINTS)

# save SIFT
cv2.imwrite("2_matches.png", img3)


# Convert the matched keypoints to coordinates
src_pts = np.float32([kp1[m.queryIdx].pt for m in good]).reshape(-1, 1, 2)
dst_pts = np.float32([kp2[m.trainIdx].pt for m in good]).reshape(-1, 1, 2)

# Homography and RANSAC
homography_matrix, inliers = RANSAC(src_pts, dst_pts, 100, 5)

# save Inliers
inlier_matches = [good[i] for i in inliers]
img_inliers = cv2.drawMatches(img1, kp1, img2, kp2, inlier_matches, None, flags = cv2.DrawMatchesFlags_NOT_DRAW_SINGLE_POINTS, matchColor = (0,255,0))
cv2.imwrite("2_inliers.png", img_inliers)

# save Outliers
outlier_matches = [good[i] for i in range(len(good)) if i not in inliers]
img_outliers = cv2.drawMatches(img1, kp1, img2, kp2, outlier_matches, None, flags = cv2.DrawMatchesFlags_NOT_DRAW_SINGLE_POINTS, matchColor = (255,0,0))
cv2.imwrite("2_outliers.png", img_outliers)

dst = cv2.warpPerspective(img1, homography_matrix, (img1.shape[1]*2 + img2.shape[1], img1.shape[0]*2), flags=cv2.INTER_CUBIC)

dst[0:img2.shape[0], 0:img2.shape[1]] = img2
# Save the result
cv2.imwrite("panorama.png", dst)