HoG Image Classifier Bitmap Foreground Masking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image classification methods using Histograms of Oriented Gradients (HoG) are computationally expensive and inefficient, particularly for mobile devices, due to the need to process entire images and handle various scales and backgrounds, leading to high power consumption and limited utility on devices like phones and tablets.
Innovation Solution
The method involves generating a non-rectangular silhouette from training images to identify relevant foreground information, reducing the HoG calculation and comparison to only these pixels, and using a bitmap to limit the processing area, thereby reducing computational costs and increasing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If HoG calculation is applied to entire images to ensure comprehensive object detection, then detection accuracy is improved, but computational expense increases significantly
Solution Approach 1:
The patent segments the image processing task by dividing the image into a grid of cells and further into blocks, where HoG calculations are performed only on relevant blocks containing foreground objects rather than the entire image. This segmentation enables selective processing that maintains detection accuracy while reducing computational load.
Solution Approach 2:
The patent extracts and processes only the relevant portions of the image containing foreground objects using bitmaps that identify active cells and blocks. By taking out and processing only these relevant regions rather than the entire image, the system achieves significant computational savings while preserving detection accuracy for objects of interest.
2Reliability
If HoG calculation is performed on all cells in the image to ensure complete coverage, then detection reliability is improved, but power consumption increases
Solution Approach 1:
The patent applies partial action by performing HoG calculations only on blocks that contain foreground objects rather than all blocks in the image. The bitmap-based selection mechanism enables the system to process only the necessary portion of the image, reducing power consumption while maintaining sufficient detection reliability for the objects of interest.
3Adaptability or versatility
If comprehensive image processing is applied to handle various scales and backgrounds, then adaptability is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary actions by generating bitmaps that identify foreground regions before executing the full HoG calculation. This preliminary segmentation allows the system to adapt to various scales and backgrounds by pre-identifying relevant regions, thereby reducing the overall processing time while maintaining adaptability to different image conditions.
Data Source
AI summary
Systems and methods are provided for image classification using histograms of oriented gradients (HoG) in conjunction with a trainer. The efficiency of the process is greatly increased by first establishing a bitmap which identifies a subset of the pixels in the HoG window as including relevant foreground information, and limiting the HoG calculation and comparison process to only the pixels included in the bitmap.