Switched-capacitor networks for scale-space generation

Abstract

In scale-space filtering signals are represented at several scales, each conveying different details of the original signal. Every new scale is the result of a smoothing operator on a former scale. In image processing, scale-space filtering is widely used in feature extractors as the Scale-Invariant Feature Transform (SIFT) algorithm. RC networks are posed as valid scale-space generators in focal-plane processing. Switchedcapacitor networks are another alternative, as different topologies and switching rate offer a great flexibility. This work examines the parallel and the bilinear implementations as two different switched-capacitor network topologies for scale-space filtering. The paper assesses the validity of both topologies as scale-space generators in focal-plane processing through object detection with the SIFT algorithm.

CMOS-3D Smart Imager Architectures for Feature Detection

“…This has two major drawbacks: on the one hand, it raises nonlinearity problems; on the other hand, it requires high precision circuits to control the time constant and hence set the sigmalevel for Gaussian filtering. On the contrary, we herein employ discrete-time resistive grids with resistors emulated by SC circuits -based on our previous proposal in [36]. Thus, sigmalevels are controlled in very simple manner, by setting the number of clock cycles, and the operation is inherently linear.…”

“…The diffusion or exchange of charge takes place with the neighbors located along the four cardinal directions with two non-overlapping clock cycles (signals φ 1 and φ 2 in Fig.7(a) by means of the bi-linear SC network displayed on Fig. 7(a) [36]. Fig.…”

“…By looking at equations (7) and (8), it is easy to identify the σ achieved per each individual cycle [36], as indicated in Eq. (10).…”

“…8. Signals φ dif f si are used for initialization purposes at the beginning of every scale [36]. Signals φ dif f oi with the ending oi meaning the octave where they are employed.…”

“…The main challenge for the extraction of Gaussian pyramids is to implement programmable Gaussian filters in accurate and controllable manner. We particularly employ SwitchedCapacitor (SC) networks, based on our previous proposal reported in [36], as this method enables to emulate inherently linear diffusion networks, which is advantageous versus using nonlinear resistors [16], [37]. Besides this, other relevant feature of the herein reported architecture arises from the possibility of in-pixel processing elements to be multiplexed in time to operate on different data and to be used for concurrent implementation of CDS (Correlated Double Sampling) and ADC (Analog-to-Digital Conversion).…”