A 3.4-$\mu$W Object-Adaptive CMOS Image Sensor With Embedded Feature Extraction Algorithm for Motion-Triggered Object-of-Interest Imaging

An Energy/Illumination-Adaptive CMOS Image Sensor With Reconfigurable Modes of Operations

“…2, analog power consumption takes a little more than 40% of total power consumption. In order to reduce the analog power consumption, many low power imagers employed in-pixel comparators for ADC [13]- [16] instead of column-parallel comparators. However, in-pixel comparators take the significant portion of pixel area and increases the pixel size.…”

“…A vision sensor that asynchronously generates outputs only when events are detected is useful for WSN applications because the amount of data can be significantly reduced [32], [33]. However, these vision sensors should include in-pixel circuits for detecting events based upon contrast or motion changes [34], which increase a form factor of the sensor and may not be able to provide post image processing capabilities such as object detection and recognition [13]. Some image sensors employed in-pixel pulse-width-modulation to achieve low power consumption from low supply voltage instead of conventional analog readout [13]- [16], [28].…”

“…However, these vision sensors should include in-pixel circuits for detecting events based upon contrast or motion changes [34], which increase a form factor of the sensor and may not be able to provide post image processing capabilities such as object detection and recognition [13]. Some image sensors employed in-pixel pulse-width-modulation to achieve low power consumption from low supply voltage instead of conventional analog readout [13]- [16], [28]. Even though these low power imagers enabled a long lifetime of the sensor node, an additional challenge for wireless sensor nodes is to accommodate the wide range of illumination.…”

Retrieving gray-level information from a Binary Sensor and its application to gesture detection

“…There exists a number of sensors that can perform some form of visual processing on chip [1,2], or extract salient features from the scene [3,4,5,6]. We modify an HDR, binary contrast sensor to enable the extraction of gray-level information.…”

“…Some rely on temporal contrast, implemented through frame differencing [6,5,7], or frame-free light intensity change [3]; others focus on spatial contrast, either using some flavor of pixellevel connectivity [8,9], or by performing spatial-temporal operations [1,10]. In our work, we focus on spatial contrast, an essential cue in the context of image and video analysis.…”