Scale-Based Image Sensor Readout for Low-Power Vision Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Computer vision processing in devices like smart glasses faces challenges due to high computational complexity and power constraints, as existing techniques require sequential row-by-row readouts and multiple image captures, leading to inefficiencies and increased power consumption.
Innovation Solution
Implementing a scale-based sensor readout with bit-reversed addressing and memory buffers to enable parallel processing at different scales, allowing early termination and reduced power consumption by skipping redundant reads and using reduced precision integral images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sequential row-by-row readout is used, then complete image data is transferred, but processing time and power consumption increase
Solution Approach 1:
The patent performs preliminary processing at reduced scales before full-resolution processing. By reading and processing image data at multiple scales in out-of-sequence manner, the system can terminate processing early when sufficient information is obtained, avoiding complete data transfer and reducing processing time while maintaining adequate measurement precision.
Solution Approach 2:
The patent divides the image data processing into multiple scales and segments. Instead of processing the entire image at full resolution sequentially, the system segments the processing into different scale levels, allowing early termination and parallel processing operations to reduce overall processing time.
2Loss of information
If full data transfer is performed, then complete image information is available, but power consumption increases
Solution Approach 1:
The patent applies partial action by transferring and processing only the necessary portion of image data at each scale level. By using variable length row reads and out-of-sequence processing, the system transfers minimal required data to achieve the processing goal, reducing power consumption while maintaining sufficient image information completeness for the application.
3Loss of substance
If variable length row reads are used, then data transfer is reduced, but readout complexity increases
Solution Approach 1:
The patent changes the readout parameters by implementing variable length row reads based on scale factors. The readout logic dynamically adjusts the number of pixels read per row based on the current scale level and bit-reversed addressing scheme, reducing data transfer volume while managing readout complexity through systematic parameter variation rather than complex ad-hoc logic.
Data Source
AI summary
Methods and apparatus for scalable processing. Conventional image sensors read image data in a sequential row-by-row manner. However, image data may be more efficiently processed at different scales. For example, computer vision processing at a first scale may be used to determine whether subsequent processing with more resolution is helpful. Various embodiments of the present disclosure readout image data according to different scales; scaled readouts may be processed using scale specific computer vision algorithms to determine next steps. In addition to scaled readouts of image data, some variants may also provide commonly used data and/or implement pre-processing steps.