Correlators architecture: a small survey

The most common types of correlators are surveyed in this post. The aim is to provide some basic concepts about correlators such as 4-f correlator (Vander Lugt correlators), Joint-transform correlator, and some others. Only 2-D correlators are surveyed.

Vander Lugt (4-f) Correlator

Such type of correlator is mainly the spatial frequency filter. To obtain the cross-correlation of g(x,y) and h(x,y) , we need to synthesize a consistent to h(x,y) filter H(u,v) .

Figure 1: 4-f correlator's scheme

PE coding scheme

Thus, multiplication of Fourier filters occurs in Vander Lught's correlators. And you need to record a hologram or interfering image of the reference image. It is pretty difficult if you want to operatively compare images. After synthesizing of corresponding correlation filter, one should set it up in the P2 plane and turn it by 180 degrees.

The intensity of correlation peaks are connected with similarity of reference image and input scene. Colier (page 557): there is a difficulty. Because of linearity of our correlator and presence of loads of correlation signals.

It is interesting to note that correlation peak wouldn't be the only one, and it's width would be different.

Joint transform correlator

Colier page 563 writes about it: JTC correlator has many practical applications. To do it best, one should use spatial light modulators.

In the JTC correlators, it is much less strict rules for positioning the reference image. It is best for real-time applications, when both input scene and reference image are shine upon the correlator.

Figure 2: JTC scheme

However, the presence of complex or expensive elements complicates the creation of simple and low-priced devices of this type.