cal 
tal 
ne 
1d 
on 
tor 
his 
ge 
ue 
tor 
od 
en 
5 
261 
3. SENSOR STRUCTURE 
3.1. Organization 
A 16x32 retina has been designed from the described pixel. This retina has been organized in 32 lines of 16 pixels. Each 
line, selected by a shift-register, is read by an interface card (acquisition and control) implemented in a PC compatible 
microcomputer. The figure 3 shows the chip organization. 
n NN A Rs BABE As wh 
^ ^ 
^o^ A A A8 à à 
^o^ ^^ * ^ » Shi ft t AT AA AA AA 
^o^ ^ I -regis er. a^ A Ah ^^ 
AA A AAA ^^^ 
AA AAA 
16x32 pixels 
sensor 
16 outputs 
  
Figure 3: Sensor's organization 
3.2 Local edge image 
In the natural image sensing conditions, spatial contrast informations can be translated into pure local temporal one by 
trembling movement of the sensor. In the figure 4, the retina has a trembling movement in front of a static lighting spatial 
edge. All the adapted pixels are those that never see the lighting spatial edge. The others detect a lighting 
variation; they are not adapted. If the retina's trembling movement is a sine movement, the evolution of the photodiode 
voltage sampled at the end of the exposure time has equally a sine form. If the derivative's amplitude of this function is 
sufficiently great, the variation is detected by the pixel. By sensing this temporal informations, spatial contrast 
informations can be extracted. Likewise, FPN problem in the neighbour pixels is avoid because the spatial information is 
in fact sensed by the same pixel. 
It's difficult to detect explicitly the pixels which receive the lighting variation in the "0" and "1" sequence. So some more 
explicit detection is needed. The principal idea is to do a sliding addition of each binary output on n periods. If the result 
is O or n, then the concerned pixel has "seen" respectively a positive or a negative lighting variation; it is on an edge. 
In the other cases the pixel is considered in his adapted state: it is not on an edge. So, a local image edge can be 
extracted by this mean. This simple processing is easily implemented with modest performance microprocessor. 
EEE Li 
Ed OPI popu 
E lil 
Ab s 
RB 
PTT TT 
  
Trembling movement of the sensor 
Figure 4: Local edge extraction by trembling movement. 
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop “From Pixels to Sequences’, Zurich, March 22-24 1995