A new, front-end image processing chip is presented for real-time small obj
ect detection. It has been implemented using a 0.6 mu, 3.3V CMOS technology
and operates on 10-bit input data at 54 megasamples per second. It occupie
s an area of 12.9mm x 13.6mm(including pads), dissipates 1.5W, has 92 I/O p
ins and is to be housed in a 160-pin ceramic quarter flat-pack. It performs
both one- and two-dimensional FIR filtering and a multilayer perceptron (M
LP) neural network function using a reconfigurable array of 21 multiplicati
on-accumulation cells which corresponds to a window size of 7 x 3. The chip
can cope with images of 2047 pixels per line and can be cascaded to cope w
ith larger window sizes. The chip performs two billion fixed point multipli
cations and additions per second.