next up previous contents
Next: A Floating-Point Convolution Up: Sensor and Processor Previous: High-Speed Miniature Orienting

A VLSI Foveated Retina in Standard CMOS

Foveation is a biologically motivated image transformation which has attracted the interest of researchers in the fields of computer vision and robotics. Its principle advantages are the realization of a high degree of image compression as well as the property of scale and rotation invariance. To date various computational implementations, as well as a fully custom CCD image sensor have been proposed. While these approaches achieve adequate performance, they nevertheless suffer from the need for considerable support resources such as networks of DSP processors and digital frame-grabbers. These resources may be readily available in a laboratory environment. However, truly autonomous mobile systems will require foveated sensors which are extremely compact and energy efficient. Recently, there has been an increase in the use of standard CMOS processes for the implementation of image sensors. Such image sensors benefit from the integration of image sensing and processing functions on the same die, yielding a reduction in power consumption and system mass. These savings make possible the realization of a completely self-contained foveated image sensor for use on mobile robots. We have designed, fabricated and tested such a device for use in a robot eye for an autonomous robot system. The fabricated prototype supports frame rates as high as 2000 frames/second, and nominally consumes 10 mW from a 3.3 V supply while compressing a 500 x 500 pixel image down to 6176 data bytes.

R. Wodnicki, G. Roberts, M.D. Levine

VLSI Foveated Sensor. On the left, the original image contains a region of detailed information at its center combined with low resolution information to highlight the Log-Polar mapping in the periphery of the sensor. On the right, the outputs of the sensor show correct reproduction of fine detail in the original image (Fovea Image), combined with data reduction due to the Log-Polar mapping (Periphery Image) levine:bobertw



next up previous contents
Next: A Floating-Point Convolution Up: Sensor and Processor Previous: High-Speed Miniature Orienting



Thierry Baron
Mon Nov 13 10:43:02 EST 1995