A 1-D CMOS digital pixel image sensor system architecture is presented. Each pixel contains a photodiode, a low-power
charge-sensitive amplifier, low noise sample/hold circuit, an 8-bit single-slope ADC, a 12-bit shift register and timing &
control logic. The pixel is laid out on a 4µm pitch to enable a cost efficient implementation of high-resolution pixel
arrays. Fixed pattern noise (FPN) is reduced by a charge-sensitive feedback amplifier, and the reset noise is cancelled by
correlated double sampling read out. A prototype chip containing 512 pixels has been fabricated in the TSMC .25um
logic process. A 40μV/e- conversion gain is measured with 100 e- rms read noise.
Image sensor sensitivity is critical for machine vision applications where illumination is limited and large depth of field is required. In this paper a method is presented for evaluating image sensor sensitivity by measuring camera signal-to-noise ratio (SNR). The method is simple to implement and produces accurate results. The method measures SNR as a function of target illumination, and relates this to image sensor sensitivity by normalizing with respect to image sensor pixel size. The method can be used to compare cameras with different types and sizes of image sensors, and is particularly useful in comparing the sensitivities of different CMOS sensors, and for comparing CMOS and CCD sensors. We present a new measure of sensitivity: SNR per unit luminous flux-time. The new measure enables the direct comparison of sensor performance.
A unified approach to image focus and defocus analysis (UFDA) was proposed recently for three-dimensional shape and focused image recovery of objects. One version of this approach which yields very accurate results is highly computationally intensive. In this paper we present a parallel implementation of this version of UFDA on the Parallel Virtual Machine (PVM). One of the most computationally intensive parts of the UFDA approach is the estimation of image data that would be recorded by a camera for a given solution for 3D shape and focused image. This computational step has to be repeated once during each iteration of the optimization algorithm. Therefore this step has been sped up by using the Parallel Virtual Machine (PVM). PVM is a software package that allows a heterogeneous network of parallel and serial computers to appear as a single concurrent computational resource. In our experimental environment PVM is installed on four UNIX workstations communicating over Ethernet to exploit parallel processing capability. Experimental results show that the communication over-head in this case is relatively low. An average of 1.92 speedup is attained by the parallel UFDA algorithm running on 2 PVM connected computers compared to the execution time of sequential processing. By applying the UFDA algorithm on 4 PVM connected machines an average of 3.44 speedup is reached. This demonstrates a practical application of PVM to 3D machine vision.