It would be easier to detect differences between the images as a whole than it would be for unique objects within the images. For exmaple you could compare the pixel values of both images and keep track of the number of pixels that have the same corresponing color, then inform the user the percentage of color data that the images share. You could then make suggestions based on the percentage like so:

Percentage of matching pixels

100% Same image
90 - 99% Same image minor discrepencies
50 - 89% Similar key objects within the image
0-49% different images with random matching colors

This could lead to a way to compare the trees. Find the smallest bounding box the fits both trees then compare. if you have above 80% similarites you could say that the trees are the same with minor differences.

This routine would be slow for large images though.