Blob detection

Description 

​​

• The LoG filter in 2D is a rotationally symmetric version of the 1D filter we worked with in the previous question (the famous “mexican hat”). Because of its shape, it makes a good blob detector (a blob is a dark patch on light background). We don’t know the size of the blobs to detect a priori, so we build a scale space by applying the filter with wider and wider σ. In practice we will approximate the LoG by a DoG. We will implement a simple (but inefficient) version where we do not downsample the image when blurring it.

​

​