Sparse Hough Transform for Image Feature Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital image processing methods, such as the Hough transform, are computationally intensive and resource-heavy, making them unsuitable for devices with limited processor and memory resources, such as mobile phones and digital cameras, when applied to large digital images with noise.
Innovation Solution
A method that pseudo-randomly selects a small percentage of pixels from the image array for processing using a parameterization transform like the Hough or Radon transform, reducing the number of pixels processed by up to 90% while maintaining feature identification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the Hough transform is applied to process all pixels in a digital image, then feature identification accuracy is maintained, but execution time and resource consumption increase significantly
Solution Approach 1:
The patent applies partial action by processing only a subset of pixels (e.g., every nth pixel or randomly selected pixels) rather than all pixels in the image. This reduces the computational load and execution time while maintaining sufficient feature identification accuracy for practical applications.
Solution Approach 2:
The patent segments the image processing task by dividing the pixel set into multiple groups and processing only selected groups. This could involve processing alternate pixels, randomly selected pixels, or pixels from specific regions, thereby reducing overall computation while preserving essential feature detection capability.
2Measurement precision
If the Hough transform is applied to process all pixels in a digital image, then feature identification accuracy is maintained, but processor and memory resources are excessively consumed
Solution Approach 1:
The patent reduces resource consumption by performing the transform on a partial set of pixels rather than the complete image. This partial processing approach maintains adequate feature detection while significantly lowering processor and memory requirements, making the technique feasible for devices with limited resources.
Solution Approach 2:
The patent extracts and processes only the necessary subset of pixels required for feature identification, rather than processing the entire image dataset. This extraction approach removes unnecessary computational overhead while preserving the essential information needed for accurate feature detection.
3Productivity
If a small percentage of pixels are processed using the Hough transform, then execution time is reduced, but the ability to identify features in noisy images may be compromised
Solution Approach 1:
The patent carefully calibrates the percentage of pixels to process (e.g., 10%, 25%, or other fractions) to achieve an optimal balance between execution speed and feature identification reliability. This partial processing approach maintains sufficient statistical information to reliably distinguish actual features from noise.
Solution Approach 2:
The patent adjusts processing parameters such as the pixel sampling rate, transform threshold values, and accumulator thresholds to compensate for processing fewer pixels. These parameter changes ensure that feature identification remains reliable even when analyzing a reduced subset of the image data.
Data Source
AI summary
A digital image includes a plurality of pixels arranged in an array. In a method of analyzing the image, some of the pixels are purposefully not processed. In particular, only those pixels in a particular subgroup are processed according to a Hough or similar transform. The number of pixels in the subgroup is less than the total number of pixels in the image (e.g., as little as about 5% of the total pixels), and each pixel in the subgroup is pseudo-randomly selected. The Hough transform is inherently configured to function within the context of noisy images, for identifying features of interest in the image, as simulated by the pseudo-random selection and processing of less than the total number of pixels in the image. This significantly reduces the processor resources required to analyze the image.