The Problem and Goal:

For monitoring traffic and detecting obstacles on highways, the intensity dynamic range of 1:10,000 is required for TV cameras. Conventional cameras, however, can provide only 1:60 with cheap cameras or 1:1,000 with cooled expensive slow cameras. The goal of this project is to develop a TV camera which has enough intensity dynamic range for ITS applications.

Previous Work:

One way to extend the intensity dynamic range is to convert the relationship between the incoming light and the camera output voltage from linear to logarithmic characteristic, as shown in Figure 1. A drawback of this approach is that intensity resolution is decreased.

Approach:

Our approach is to dynamically change the relationship between the incoming light and the camera output voltage from frame to frame, depending on the required dynamic range. When the required dynamic range is small, the input-output relationship becomes closer to the conventional linear scheme for better intensity resolution. The logarithmic characteristic is increased as the required dynamic range increases at a cost of intensity resolution.

For the brightness adaptability, a time-varying energy barrier gate and the drain were added to each pixel, as shown in Figure 2. Figure 3 describes that the height of the energy barrier, between the pixel well and the drain, is low in the early stage (TE) of each image acquisition period and is increased to a high level for the late stage (TL) of each period. At any given time, the electrons in the pixel well, with energies which exceed the gate height, are discharged over the gate and into the drain.

The output for low intensity light L1 does not change with this time-varying gate height. The output of the high-intensity light, in contrast, is suppressed from H1 to H2 because some electrons are discharged in the early stage of the acquisition period. The gate-height function can be changed dynamically based on the nature of the images. Adding more steps to the energy barrier function can produce even greater compression.

Current Status and Future Work:

The first version of 256x256-pixels camera has been developed. This version has two additional features: (1) CMOS technology was used instead of conventional CCD, and (2) parallel 10-bit analog-to-digital converters were included in the imager chip. The pixel size is 22x22 microns. Real-time format converter is under development to provide 30 frames per second. Figure 5 shows that the intensity dynamic range can be extended over 1:10,000 ratio with the compressive mode.