object detection, image matching for small objects

I am using SIFT feature the matching between two images or it can be considered as object detection.

First matching. I used SIFT feature for matching and it failed when the object is small.

Second matching. However, it woks when I cropped that object out from that image.

Any suggestions or ideas to improve the first matching without cropping?

Example images are shown below.

Image A:

Image B:

Image C cropped out from image B:

Answer 1

[Image Analyst]

This is exactly what this group at the University of Arizona did using Hu's moments: http://www.youtube.com/watch?v=Nc06tlZAv_Q Maybe you might want to try that method. The advantage is that it's robust to scale and rotation changes (within reason).