In the description of how to compute the panorama (Section XXXXXXXXXX), it is suggested that we take each pixel in the central image and transform it into the other images and then copy the color....

In the description of how to compute the panorama (Section 15.7.2), it is suggested that we take each pixel in the central image and transform it into the other images and then copy the color. What is wrong with the alternate strategy of taking each pixel from the other images and transforming them into the central image?