Multi-Resolution Co-occurrence Histogram for Robust Image Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image recognition technologies, such as those using Histograms of Oriented Gradients (HOG) feature amounts, are not robust enough for effectively recognizing and tracking individuals, particularly in scenarios like in-vehicle camera surveillance where variations in image quality and resolution can impact detection accuracy.
Innovation Solution
An image feature amount output device that acquires co-occurrence in luminance gradient directions across different resolutions of an image, generating a co-occurrence matrix and histogram to serve as a more robust feature amount for image recognition, allowing for improved detection and tracking of objects like pedestrians.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If HOG feature amount is used for image recognition, then the recognition process can be performed with simple gradient direction distribution, but the robustness of recognition deteriorates under varying image quality and resolution conditions
Solution Approach 1:
The patent transitions from analyzing gradient directions at a single resolution to analyzing co-occurrence of gradient directions across multiple resolutions. This dimensional expansion from 1D (single resolution) to 2D (multiple resolutions) allows the system to capture spatial relationships that are invariant to scale changes, thereby improving robustness while maintaining computational feasibility through histogram-based aggregation
2Reliability
If multiple resolutions are analyzed to improve robustness, then the reliability of image recognition is improved, but the complexity of the processing system increases
Solution Approach 1:
The patent divides the image into multiple resolution levels (e.g., original resolution and downsampled versions) and processes each level separately to extract gradient direction histograms. This segmentation approach allows the system to capture features at different scales independently, then combines them through co-occurrence analysis to achieve robustness without requiring complex unified processing models
3Reliability
If co-occurrence of gradient directions across multiple resolutions is calculated, then the robustness of image recognition is improved, but the computational time and processing complexity increase
Solution Approach 1:
The patent pre-calculates and stores gradient direction histograms for each resolution level before performing co-occurrence analysis. By preparing these intermediate results in advance (preliminary action), the system avoids redundant calculations during the actual recognition process, significantly reducing computational time when evaluating multiple resolutions and their co-occurrences
Data Source
AI summary
An image processing device converts an image that is a recognition object image to high-resolution, medium-resolution, and low-resolution images. The device sets the pixel of interest of the high-resolution image, and votes the co-occurrence in a gradient direction with offset pixels, the co-occurrence in the gradient direction pixels in the medium-resolution image, and the co-occurrence in the gradient direction pixels in the low-resolution image, to a co-occurrence matrix. The device creates such a co-occurrence matrix for each pixel combination and for each resolution. The device executes the process on each of the pixels of the high-resolution image, and creates a co-occurrence histogram wherein the elements of a plurality of co-occurrence matrices are arranged in a line. The device normalizes the co-occurrence histogram and extracts, as a feature quantity of the image, a vector quantity having as a component a frequency resulting from the normalization.