Biologically-based Stereopsis: Theories and VLSI Implementation
Stereopsis is an important method used by humans to perform depth perception. Each eye, being separated in space, receives a slightly different view of the world. Stereopsis is the procedure that utilizes the differences in views to determine the relative depth of objects. The goal of this work is to develop an analog VLSI opto-electronic system that can perform stereopsis. Various steropsis algorithms and models are examined, with an eye toward analog VLSI implementation. Based on the information obtained from these methods, a novel stereopsis algorithm is introduced. This algorithm has been developed so that it performs stereopsis and is suited to implementation in silicon VLSI. The algorithm has been tested via a software simulation that uses various scenes (pictures) as inputs. An image library has been developed which can be used as a standard testbed of images for comparing stereopsis processors. From the software model, a circuit model has been developed. The stereopsis processor is an array of cells with each cell composed of a replica of the stereopsis circuit model. The system is based on a clocked-analog method to increase the amount of processing that can be done in a smaller area. The system is developed in stages to simplify testing and analysis. The final system is a two-dimensional array with optical input that produces a voltage output corresponding to a particular disparity. This analog VLSI system represents the next step in the development of artificial visual systems that mimic the functionality of vertebrates’ visual systems.
Back