Data rate control for event-based vision sensor

Abstract

To provide a dynamic vision sensor (DVS) or a change detection sensor configured to control or moderate event rate.SOLUTION: A dynamic vision sensor (DVS) or change detection sensor is controlled to keep an event rate close to a desired rate or within desired bounds. Adapting the configuration of the sensor to the scene by changing ON-event and / or OFF-event thresholds, allows necessary amount of data, but not much more than necessary to be obtained such that the overall system gets as much information about its state as possible.SELECTED DRAWING: Figure 1A

Description

[Technical Field]

[0001] [Related Applications] This application claims priority to Swiss Provisional Patent Application No. CH20160001764, filed December 30, 2016, and Swiss Provisional Patent Application No. CH20160001765, filed December 30, 2016, both of which are incorporated herein by reference in their entirety. [Background technology]

[0002] Today, machine vision is largely based on conventional cameras and their associated frame-based image sensors. For some machine vision tasks, such as object recognition, these conventional frame-based cameras are well suited. However, for other tasks, such as tracking or position and motion estimation, conventional image sensors have deficiencies.

[0003] A major drawback is that conventional cameras generate a significant amount of redundant and unnecessary data that must be captured, communicated, and processed. This high data workload slows reaction time by reducing temporal resolution, leads to increased power consumption, and increases the size and cost of machine vision systems. Furthermore, most image sensors suffer from limited dynamic range, poor low-light performance, and motion blur.

[0004] These imperfections arise from the fact that the data is captured as a series of still images (frames). In some cases, encoding dynamic scenes as still images is useful for producing beautiful images and videos without data processing optimization, but this is not very important for many machine vision applications.

[0005] Conventional computer vision systems using conventional cameras typically compare features between successive image frames for object recognition. Two successive overlapping images taken at different times and from different poses are compared to estimate the position and orientation of a moving system and to infer a three-dimensional map of the surrounding world. To infer the movement that occurred between the two frames, characteristic landmarks (keypoints or other visual features) must be matched across the two images. Finding these pairs of corresponding points in both images is known as solving the "correspondence problem."

[0006] Solving the correspondence problem requires a significant amount of processing power. To detect landmarks, every pixel in an image may need to be searched for distinctive features (corners, blobs, edges, etc.). The pixels and their surrounding neighboring pixels are then grouped to characterize what are called feature descriptors, which are used to match features between frames, thereby establishing pairs of corresponding points. This is computationally intensive. Direct approaches that directly compare pixel intensities are even more computationally complex.

[0007] On the other hand, so-called dynamic vision sensors (DVS) or event-based change detection sensors overcome the limitations of frame-based coding. By using intra-pixel data compression, data redundancy is eliminated, achieving high temporal resolution, low latency, low power consumption, and high dynamic range with little motion blur. Therefore, DVS is particularly well suited for solar- or battery-powered compressed sensing, or for mobile machine vision applications where the system's position needs to be estimated and processing power is constrained by limited battery capacity.

[0008] A DVS locally preprocesses visual information. Instead of generating a sharp image, a DVS generates smart data for computer applications. While traditional image sensors capture a movie as a series of still images, a DVS simply detects and communicates the location of changes in a scene. It therefore encodes visual information much more efficiently than traditional cameras for intra-pixel data compression. Specifically, rather than encoding frames as image data, a DVS detects changes and encodes those changes as change events. This means that data can be processed using fewer resources, lower net power, and with faster system response times. High temporal resolution allows for continuous tracking of visual features, thereby overcoming the correspondence problem. Furthermore, the DVS architecture enables high dynamic range and good low-light performance. Summary of the Invention [Means for solving the problem]

[0009] A common characteristic of DVS or change detection sensors is that the output data rate is variable because the rate of events depends on the scene the sensor is observing and the sensor configuration.

[0010] A variable amount of change events being read from the sensor can cause problems on the sensor and at the system level, especially if more events are being read than the processor can process in real time.

[0011] Discarding an event when it cannot be processed results in data loss, and therefore the system loses information about its state (e.g., about its motion and position). In the case of a change detection sensor, the result can be that the sensor misses events that occur in, for example, a portion of its pixel field.

[0012] Adapting the sensor configuration to the scene allows the overall system to have the necessary, but not too much, data to obtain as much information as possible about its state. On the other hand, the number of events can be controlled based on the state of the sensors. For example, to reduce power, it may be desirable to reduce the number of events, such as during sleep cycles, and then increase the number of events during normal operating mode.

[0013] The system proposed herein alters the configuration of the chip or sensor to control or regulate the event rate, for example, this control can be used to keep the event rate near a desired rate or within a desired range.

[0014] Nevertheless, the amount of events generated during a given time slice can only be conclusively determined after outputting and counting them, where some embodiments use an event count estimator that allows a rough estimation of the number of pending events before reading them out.

[0015] With even a rough estimate of the number of pending events, the system can determine whether it can cope with the data and read them, or whether there are too many (or too few) events, and the system can redo the change detection using a higher threshold to get fewer events (or redo the comparison with a lower threshold to get more events).

[0016] In general, according to one aspect, the invention features a change detection sensor. The sensor includes a pixel array including pixels that detect light, and an event detector for detecting events associated with the light received by the pixels. In accordance with the invention, an event rate detector is provided for evaluating the events. A controller then changes how the pixels detect the events based on the event rate detector.

[0017] In one embodiment, the event rate detector includes a counter for evaluating the events by counting the events from the event detector.

[0018] In another embodiment, the event rate detector comprises an event estimator for evaluating the events by estimating the events, which may be performed by analyzing an analog signal based on the number of events recorded by the event detector.

[0019] Preferably, the sensor includes a threshold generation circuit for setting the threshold applied by the event detector, the controller then varying the threshold provided in response to the assessment provided by the event rate detector.

[0020] Preferably, the controller sets thresholds for ON and OFF events separately based on the ON and OFF events in the pixel array.

[0021] In the current embodiment, each of the pixels includes a memory capacitor in one of the event detectors for storing a charge corresponding to the light it receives when the pixel is reset, and each of the pixels may further include one or more comparators in the event detector for comparing changes in the received light with one or more thresholds for detecting an event.

[0022] In another embodiment, the comparator is located within the readout circuit.

[0023] In general, according to another aspect, the invention features a method of operation of a change detection sensor that includes detecting an event associated with light received by a pixel in a pixel array, evaluating the event from the pixel array, and, based on the evaluation of the event from the pixel array, changing how the pixel detects the event.

[0024] The above and other features and advantages of the present invention, including various novel details of construction and combinations of parts, will now be more particularly described with reference to the accompanying drawings and pointed out in the claims. It is to be understood that the particular methods and devices embodying the invention are shown by way of illustration and not as limitations of the invention. The principles and features of this invention can be employed in a wide variety of embodiments without departing from the scope of the invention.

[0025] In the accompanying drawings, reference characters refer to the same parts throughout the different views. The drawings are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the invention. [Brief explanation of the drawings]

[0026] [Figure 1A] FIG. 1 is a block diagram of an event-based change detection sensor including an event rate control. [Figure 1B] FIG. 1 is a circuit diagram illustrating an example of a change detection pixel that can be used in pixel array 110. [Figure 2] FIG. 2 is a block diagram illustrating an embodiment of an event rate detector implemented as a periodically reset event counter. [Figure 3A] 10 is a flow diagram illustrating the operation of the controller 120 for controlling the threshold generation circuitry based on the event rate detector. [Figure 3B] 10 is a flow diagram illustrating the operation of the controller 120 for controlling the threshold generation circuitry based on the event rate detector. [Figure 4] FIG. 10 is a schematic diagram illustrating an event-based change detection sensor that generates an event rate estimate according to another embodiment. [Figure 5] FIG. 10 is a schematic diagram illustrating an event-based change detection sensor for generating an event rate estimate according to yet another embodiment. [Figure 6] FIG. 10 is a circuit diagram illustrating a possible implementation of the current ADC 212 of the previous embodiment. [Figure 7] 1 is a state diagram illustrating a possible implementation of controller 120. FIG. [Figure 8] FIG. 1 is a block diagram showing a basic threshold generation circuit. DETAILED DESCRIPTION OF THE INVENTION

[0027] FIG. 1A is a block diagram of an event-based change detection sensor 100 with event rate control.

[0028] In general, the change detection sensor 100 comprises five elements that are particularly relevant to current systems: an array 110 of change detection pixels; an event rate detector 200; a controller 120; a threshold generation circuit 130; a read circuit 140; Includes.

[0029] The pixel array 110 generates a variable amount of data, the change events. The event rate detector 200 periodically evaluates the events, for example, by counting or estimating the change events. The controller 120 then adapts the sensor configuration so that in the next time step, the pixel array 110 generates an amount of change events that more closely matches the desired rate.

[0030] The pixel array may be a change detection pixel such as that described in U.S. Patent Application Publication No. 2008 / 00135731 by Lichtsteiner et al., entitled "Photoarray for Detecting Time-Dependent Image Data," which is incorporated herein by reference; or a change detection pixel such as that described in U.S. Patent Application No. 15 / 858,427, filed December 29, 2017, entitled "Dynamic Vision Sensor Architecture" (hereinafter the Berner patent application, which is incorporated herein by reference in its entirety); or a change detection pixel such as that disclosed in J. Kramer, "An on / off transient imager with event-driven, asynchronous readout," IEEE International Symposium on Circuits and Systems, 2002, Vol. 2, pp. 165-168, 2002; or alternatively, a change detection pixel such as that disclosed in Ruedit et al., "A 128x128 pixel 120-dB dynamic-range vision-sensor chip for image contrast." and orientation extraction,” IEEE Journal of Solid-State Circuits, Vol. 38, No. 12, pp. 2325-2333, December 2003, or any array of spatial contrast detection pixels.

[0031] FIG. 1B is a circuit diagram of one possible change detection pixel 112.

[0032] This exemplary pixel 112 of the two-dimensional array 110 includes a photosensor, such as a photodiode (PD), that detects or measures impinging light and converts the light intensity into a signal, where the signal is a current, Iphoto. The pixel 112 also includes an event detector 114, which determines whether there is a sufficient change in the light detected by the photosensor, such that an event has occurred. This is accomplished by monitoring the current, Iphoto.

[0033] More specifically, the exemplary event detector 114 includes a photoreceptor circuit PRC, which in turn generates a photoreceptor signal Vpr in response to light intensity. A memory capacitor C1 then stores past photoreceptor signals.

[0034] Comparator A1 compares the difference between the current photoreceptor signal Vpr and the past photoreceptor signal with a threshold Vb, which is provided by threshold generation circuit 130.

[0035] Preferably, the comparator A1 or pair of comparators detects ON events based on an ON event threshold and detects OFF events based on an OFF event threshold, where the ON event is characteristic of an increase in light received by the light sensor and the OFF event is characteristic of a decrease in light received by the light sensor. If a single comparator is used, the ON event threshold and the OFF event threshold are provided successively in time as threshold Vb.

[0036] This comparator A1 can reside within each pixel or can be shared among a subset of pixels (e.g., columns). In one case, the comparator is shared and located within readout circuitry 140. Various examples of such configurations are disclosed in the Berner patents.

[0037] However, in a preferred embodiment, the comparators A1 will be incorporated into the pixels 112, with each pixel having one or more dedicated comparators A1.

[0038] The memory 50 stores an event or events. The event is taken from the comparator output Vcomp based on a sample signal from the controller 120. The memory can be a sampling circuit (e.g., a switch and parasitic or explicit capacitor) or a digital memory circuit (latch or flip-flop). In one embodiment, the memory is a sampling circuit, and each pixel has two memories, one to store any ON events and one to store any OFF events. The memory provides an output on one or more lines Evt.

[0039] The conditional reset circuit R1 provides a conditional reset based on a combination of the stored state of the comparator output and a reset signal GlobalReset applied by the controller 120.

[0040] The pixel circuit 112 and controller 120 operate as follows.

[0041] Changes in light intensity received by photosensor PD will translate into changes in photoreceptor signal Vpr. When reset circuit R1 is in a non-conducting state, changes in Vpr will also be reflected in the voltage Vdiff at the inverting input (-) to comparator A1. This occurs because the voltage across memory capacitor C1 remains constant.

[0042] At a time selected by the controller 120, the comparator A1 of the event detector 114 compares the voltage (VDiff) at the second terminal of the memory capacitor C1 with the threshold voltage Vb (from the threshold generation circuit 130) applied to the non-inverting input (+) of the comparator A1.

[0043] The controller 120 operates the memory 50 to store the comparator output Vcomp. The memory 50 is typically implemented as part of the pixel circuit 112 as shown.

[0044] However, in other embodiments, memory 50 is implemented as part of readout circuitry 140 (one peripheral circuit for each column of pixel array 110).

[0045] In the illustrated embodiment, the sensor's readout circuitry 140 reads out the memory 50 of each pixel 112 in the array 110. In one example, the readout circuitry 140 stores the coordinates of the pixel in the array that detected the change event.

[0046] If the stored state of the comparator output held in memory 50 indicates a sufficient change in light intensity (i.e., an event) and the global reset signal GlobalRest from controller 120 is active (AND), conditional reset circuit R1 becomes conductive, where "AND" denotes a logical AND operator. With conditional reset circuit R1 in a conductive state, the voltage at the inverting input of comparator A1 (Vdiff) is reset to a known level. This therefore stores the current photoreceptor signal Vpr on memory capacitor C1.

[0047] Event counting

[0048] FIG. 2 shows one embodiment of an event rate counter.

[0049] Here, event rate detector 200 is implemented as a counter 210 that is periodically reset by controller 120. The counter is incremented based on events read by readout circuit 140. The count of events is provided to controller 120. Counter 210 is then reset by controller 120 when a new count is required.

[0050] In one example, the event rate detector 200 counts the number of events output by the sensor array 110 during a given time window, where this is accomplished by the readout circuit sending an increment signal to the counter 210 each time it reads an event from the pixel array 110.

[0051] In one particular example, event rate detector 200 counts both ON events output by sensor array 110 and OFF events output by sensor array 110 during a given time window. This is preferably accomplished using separate ON event counters and OFF event counters in counter 210.

[0052] FIG. 3A is a flow diagram illustrating the operation of the controller 120 for controlling the threshold generation circuit 130 based on the event rate detector 200, according to one embodiment.

[0053] Specifically, in step 310, the counter 210 of the event rate detector 200 is reset by the controller 120 when a new count is required.

[0054] In step 312, the controller 120 reads the event count after a predetermined time.

[0055] Next, in step 314, the controller calculates a change in the event threshold to be applied to the pixel array 110. This new threshold is sent to the threshold generation circuit 130 in step 316. This threshold is then applied to the event detectors 114 of the sensor array 110 as the threshold Vb in determining the next set of events.

[0056] FIG. 3B is a flow diagram illustrating the operation of the controller 120 for controlling the event rate detector 200 according to another embodiment.

[0057] In this embodiment, the controller 120 waits for an external trigger in step 318 before resetting the counter in step 310. This embodiment may be most useful in situations where the sensor needs to be synchronized to an external timing source.

[0058] Event Count Estimator

[0059] The second approach relies on generating an event rate estimate. This approach uses an event count estimator internal to event rate detector 200. This avoids the need to read every event from the array to determine the number of events. Instead, this approach allows for a rough estimate of the number of events before the events are read out or as part of the readout process.

[0060] FIG. 4 is a schematic diagram illustrating an event-based change detection sensor 100 for generating an event rate estimate according to another embodiment.

[0061] More specifically, an event detector 114 is located in each pixel 112 of the array 110 along with a photoreceptor 116 for detecting incident light and is part of each pixel circuit. A current source 118 of each pixel circuit 112 is activated only if an event is present in that pixel.

[0062] Specifically, when the event detector 114 registers a change in the light received by the photoreceptor 116, for example due to an ON event and / or an OFF event, the event signal Evt becomes active, which closes the switch 117 associated with the pixel 112.

[0063] The currents of all pixels 112 in the array 110, or a sampling of pixels distributed across the extent of the array 110 in the sensor 100, are summed and provided to a current analog to digital converter (ADC) 212 in the event rate detector 200. This ADC 212 captures the current and converts the magnitude of the current into a digital representation. This digital representation of the sum of the currents is then provided to the controller 120 as an estimate of the number of events detected, which, if taken over a known period of time, indicates the event rate.

[0064] In general, the current sources 118 will not be very well matched between the pixels 112. This means that the sum of the currents will not allow an accurate estimation of how many pixels have pending events, but will allow a rough estimate.

[0065] FIG. 5 is a schematic diagram illustrating a change detection sensor for generating an event velocity estimate according to another embodiment.

[0066] More specifically, the photoreceptor 116 of each pixel 112 generates a signal that is sensitive to and indicative of the instantaneous amount of light detected by that photoreceptor 116. This information is sent to an event detector 114.

[0067] Thus, the output of the comparator in the event detector 114 changes when the light received by the photoreceptors 116 changes by an amount greater than the last reset threshold. Often, an ON event threshold is provided to the pixel as a pixel threshold to identify an ON event associated with an increase in the amount of light detected by the photoreceptors 116. An OFF event threshold is then provided to the pixel as a pixel threshold to identify an OFF event associated with a decrease in the amount of light detected by the photoreceptors.

[0068] In other cases, the event detector includes two comparators: a first comparator that receives an ON event threshold for identifying an ON event, and a second comparator that receives an OFF event threshold for identifying an OFF event.

[0069] When the signal Enable evt counter is active (and therefore Disable evt counter is inactive), all readout lines are shorted together by closing switch 188, and the readout lines are disconnected from the column readout circuit 180 by opening switch 190. When all readout lines are connected to the current ADC 212, current will flow from the current ADC 212 to all pixels, causing the comparator output to go high in those pixels and Evt to be active, thereby closing switch 117 in each pixel 112. Furthermore, the OR gate 186 in each row will close switch 178 in each pixel 112 because Enable evt counter is active. Therefore, the total current will depend on the number of pixels for which the output of comparator 174 is high or Evt on the signal line(s) is active.

[0070] When the signal Enable evt counter is inactive (and therefore Disable evt counter is active), the pixel readout lines are connected to their corresponding column readout circuits 180 if the corresponding Readout row select X is active.

[0071] FIG. 6 shows a possible implementation of the current ADC 212 of the previous embodiment.

[0072] The summed current is fed to an operational amplifier A214 by resistor R216 in a negative feedback configuration. The voltage difference between the reference voltage and the voltage at the output of the operational amplifier is proportional to the magnitude of the input current and therefore represents the number of pending events.

[0073] The voltage at the output of the operational amplifier is then fed to a conventional voltage ADC 218, which may be implemented as a flash ADC, a SAR ADC, a sigma-delta ADC, or the like.

[0074] After the current to voltage conversion, a voltage sampling step (between the output of OpAmp 214 and voltage ADC 218) can be added so that a potentially lengthy ADC conversion can occur while the pixel is being read out. Once the ADC conversion is complete, the pixel readout can be stopped, if desired.

[0075] Controller 120

[0076] The controller 120 is most likely a digital block with inputs for the event rate detector 200, the desired event rate (or range of desired event rates) from outside the sensor 100, and optionally a trigger input. The controller 120 outputs control signals for chip configuration (i.e., time resolution and thresholds for the threshold generation circuit 130).

[0077] The controller 120 can be implemented as a traditional P, PI or PID controller, or more simply as a state machine.

[0078] However, when using an event rate estimator, controller 120 can also be implemented as an analog circuit. The input to this analog controller is therefore not a digital representation; instead, the output voltage of operational amplifier 214 in FIG. 6 is typically fed directly to controller 120.

[0079] FIG. 7 is a state diagram of the controller 120.

[0080] The event threshold is the signal from the event rate estimator, Event rate: input from the event rate detector 200; MinEventRate: constant, minimum desired event rate, MaxEventRate: constant, maximum desired event rate, MinEventThreshold: parameter, minimum threshold setting, MaxEventThreshold: parameter, maximum threshold setting, Step: The parameter, the step at which the event threshold changes during each pass through the loop. EventThreshold: the output that is fed to the event generator; Trigger: A periodic input signal that starts the loop. is modified to keep the event rate within a desired range based on the

[0081] More specifically, the controller 120 waits for a trigger in state 710. Upon receiving it, the controller 120 compares the event rate provided by the event rate detector 200 with a minimum event rate in state 712. If the event rate is less than the minimum event rate, the controller proceeds to state 716. In state 716, the controller 120 compares the event threshold with a minimum event threshold, and if the event threshold is less, returns to the wait state at 710. On the other hand, if the event threshold is greater than the minimum event threshold, the event threshold is decreased in state 718 and the controller returns to the wait state 710.

[0082] On the other hand, if the event rate is greater than the minimum event threshold in state 712, the event rate is compared to the maximum event rate in state 714, and if the event rate is less, the controller 120 returns to the wait state 710. On the other hand, if the event rate is greater than the maximum event rate, the event threshold is compared to the maximum event threshold in state 720. If the event threshold is greater than the maximum event threshold in state 720, the controller 120 returns to the wait state 710. On the other hand, if the event threshold is less than the maximum event threshold, the event threshold is increased by one step in state 722 and the controller 120 returns to the wait state 710.

[0083] Threshold Generation Circuit

[0084] For change detection pixels such as those disclosed in U.S. Patent Application Publication No. 2008 / 00135731 by Lichtsteiner et al., entitled "Photoarray for Detecting Time-Dependent Image Data," which is incorporated herein by reference, the event threshold is the difference between the threshold voltage and the amplifier bias voltage. The On threshold voltage is typically higher than the bias voltage, while the Off threshold voltage is lower.

[0085] Basic Threshold Generation

[0086] FIG. 8 shows a basic threshold generation circuit for a pair of voltage digital-to-analog converters (DACs) or current DACs 810, 812, as disclosed, for example, in P. Lichtsteiner, C. Posch, and T. Delbruck, “A 128x128 120dB 15μs Latency Asynchronous Temporal Contrast Vision Sensor,” IEEE J. Solid-State Circuits, Vol. 43, No. 2, pp. 566-576, February 2008, and T. Delbruck, R. Berner, P. Lichtsteiner, and C. Dualibe, “32-bit Configurable bias current generator with sub-off-current capability,” Proceedings of 2010 IEEE International Symposium on Circuits and Systems, 2010, pp. 1647-1650. Specifically, for a set of DACs 810, 812, 814, there is one 814 for generating the comparator bias voltage, one 810 for the OnThres voltage, and one 812 for the OffThres voltage.

[0087] Event Rate and Balance Control

[0088] In yet another embodiment, the sensor 100 counts ON (increasing light intensity) and OFF (decreasing light intensity) events separately. The ratio between ON and OFF events is then used to control the ON and OFF thresholds separately. Generally, similar amounts of ON and OFF events over a large (greater than a second) time window are desired.

[0089] Non-motion threshold

[0090] A characteristic of event-based change detection sensors is that they generate data (or very little data) when there is no motion in the scene. In such a case, it may not be desirable to change the threshold; therefore, the controller can be extended to include a minimum event rate threshold below which the control loop for changing the EventThreshold output is not activated.

[0091] Time resolution control

[0092] In some pixel designs, the time bin rate (speed of comparison) can be varied. In pixels such as those proposed by the description in U.S. Patent Application No. US2008 / 00135731, the refractory bias can be controlled, which makes it possible to limit the rate at which the pixel can generate events.

[0093] While the present invention has been particularly shown and described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and detail can be made therein without departing from the scope of the invention encompassed by the appended claims.

Claims

1. a pixel array having pixels that detect light; an event detector for detecting an event associated with light received by the pixel; an event rate detector for evaluating the event; a controller for varying how the pixels detect the events based on the event rate detector; A change detection sensor comprising:

2. The sensor of claim 1 , wherein the event rate detector comprises a counter for evaluating the events by counting the events from the event detector.

3. The sensor of claim 1 , wherein the event rate detector comprises an event estimator for evaluating the event by estimating the event.

4. The sensor of claim 1 , wherein the event rate detector evaluates the event by analyzing an analog signal based on the number of the event detectors that register the event.

5. The sensor of claim 1 further comprising a threshold generation circuit for setting a threshold applied by the event detector.

6. The sensor of claim 1 , wherein the controller varies the threshold provided in response to the evaluation provided by the event rate detector.

7. The sensor of claim 1 , wherein the controller sets thresholds for ON and OFF events separately based on ON and OFF events in the pixel array.

8. 2. The sensor of claim 1, wherein each of the pixels comprises a memory capacitor of one of the event detectors for storing a charge corresponding to light received when the pixel is reset.

9. 10. The sensor of claim 1, wherein each of the pixels comprises one or more comparators of one of the event detectors for comparing a change in received light with one or more thresholds for detecting the event.

10. 10. The sensor of claim 1, further comprising a comparator in the event detector for comparing a change in received light with one or more thresholds for detecting the event, the comparator being located in a readout circuit.

11. 1. A method of operation of a change detection sensor, comprising: Detecting events associated with light received by pixels in the pixel array; evaluating the events from the pixel array; based on the evaluation of the event from the pixel array, varying how the pixels detect the event; A method comprising:

12. The method of claim 11 , wherein evaluating the events comprises counting the events.

13. The method of claim 11 , wherein evaluating the event comprises estimating the event.

14. The method of claim 11 , further comprising setting a threshold applied by an event detector to the pixel in response to the evaluation of the event.

15. The method of claim 11 , further comprising setting thresholds for ON and OFF events separately based on ON and OFF events from the pixel array.

Images