Project 2   Xing Li

When stitching images together, we need to registrate them first. In this project, we are using feature based method to align images. Therefore it’s very important to detect and match features accurately. The better the feature match accuracy, the better registration we can get.

According to our experience in project 1, the accuracy of feature detection and match depends on the variance between adjacent pictures.  Accordingly, when we take pictures with the tripod, it’s comparatively easy to align adjacent pictures than in the situation when we don’t use a tripod. As we can see, the first panorama, which is made from the tripod-held picture sequences, is much better than the second panorama, which is from hand-hold picture sequences.

In the project, I have tried to compensate for the brightness fluctuations while normalizing the accumulated panoramic image. The basic idea is as follow:

(1) Compute the average brightness value of the whole image by using each pixel’s RGB value.  

      Luminance = 0.30 R + 0.59 G + 0.11 B.

(2) Then the brightness of each pixel is compared with the average value and adjusted accordingly.

It seems the effect is not very clear after implementing this method. I think it is because the sun light was changing minute by minute when I took the photos ( here I an referring to the second panorama), therefore the shadows kept changing. Even I try to adjust the brightness level, I can’t change the color. So those shadows are still there.

This panorama is from the tripod-held sequence.

Click here to enjoy the Panorama.

This panorama is from the hand-held sequence.

Click here to enjoy the Panorama.

This panorama is from the test images.

Click here to enjoy the Panorama.