Contingency Table Stochastic Computing for Image Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Stochastic computing systems face challenges in simulating image processing tasks due to high latency and memory complexity, especially when dealing with long bit-streams, which are necessary for accurate computations, and current methods are inefficient in processing high-density image data.
Innovation Solution
The method employs contingency tables (CTs) to simulate stochastic computing circuits, allowing for latency-free and memory-aware emulation of image processing tasks like template matching, image compositing, and bilinear interpolation by processing scalar values directly, rather than generating and processing bit-streams, using a 2-to-1 or 4-to-1 multiplexer and XOR gate operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional bit-stream processing methods are used for stochastic computing image processing, then computation accuracy is maintained, but execution time and memory usage become excessively high
Solution Approach 1:
The patent creates a contingency table that copies and represents the statistical properties of long bit-streams in a compressed scalar format. Instead of processing actual bit-streams, the system uses a contingency table (a compact data structure) that captures the essential correlation and distribution information, enabling accurate image processing computations with dramatically reduced execution time while maintaining computational accuracy
Solution Approach 2:
The patent transforms the representation parameters from individual bit-stream sequences to aggregated statistical parameters (contingency table entries). By changing from time-domain bit sequences to frequency-domain statistical descriptors, the system achieves the same computational results with far fewer operations, resolving the contradiction between accuracy and execution time
2Measurement precision
If long bit-streams are generated for accurate stochastic computing, then computation precision improves, but memory complexity and processing overhead increase significantly
Solution Approach 1:
The contingency table serves as a compressed copy of the bit-stream statistical properties. Instead of storing and processing actual bit-stream data in memory, the system stores a compact contingency table that reproduces the same computational outcomes, dramatically reducing memory requirements while maintaining computation precision
Solution Approach 2:
The patent extracts only the essential statistical features (contingency table entries representing joint probabilities) from the full bit-stream data. This extraction eliminates the need to retain large volumes of raw bit-stream data in memory, achieving precision with minimal memory footprint by keeping only the critical computational parameters
3Adaptability or versatility
If stochastic computing systems process high-density image data, then image processing capability is enhanced, but latency and computational overhead increase
Solution Approach 1:
The patent changes the processing parameters from individual bit operations to aggregated contingency table operations. By transforming image processing computations to operate on pre-computed statistical parameters rather than raw bit-streams, the system enhances image processing capability while minimizing latency through reduced computational steps
Data Source
AI summary
Disclosed herein a method for agile simulation of a stochastic computing image processing where the input operands are processed with the aid of a correlation-controlled contingency table (CT) construct without using actual stochastic bit-streams. The disclosed method utilizes contingency tables to perform (i) template matching, (ii) image compositing, and (iii) pattern detection. Results show that the proposed approach achieves similar computation accuracy to the traditional stochastic computing simulation while performing runtime- and memory-efficient computations.