Multi-resolution Color Vector for Image Search Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image search systems are inefficient and inaccurate in finding images with similar color distributions to a query image, as they require large computational resources, introduce processing delays, and fail to capture the color distribution with sufficient detail or precision, especially when searching large numbers of images.
Innovation Solution
An image search system that determines a color vector for a query image based on a multi-resolution color histogram, applies a square root mapping to prevent dominant colors from overwhelming sparse colors, and compresses the color vector to reduce storage and computational requirements, allowing for efficient computation of distance measures between the query image and candidate images using L2-norms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional image search systems extract the most relevant colors of the query image and search for images using the relevant colors, then the search process is simplified and faster, but the precision is insufficient to find images that match the color distribution of most query images
Solution Approach 1:
The patent divides the color histogram into multiple resolutions (e.g., coarse and fine resolutions) to capture color distribution at different levels of detail. The coarse resolution provides an overview for fast comparison, while the fine resolution captures detailed color distribution characteristics, thereby resolving the contradiction between search speed and precision.
Solution Approach 2:
The patent extends the color representation from a single resolution to multiple resolutions, adding a dimensional aspect to color histogram comparison. This multi-resolution approach enables the system to simultaneously consider both broad color categories and specific color nuances, improving precision without sacrificing search efficiency.
2Measurement precision
If the image search system recolor images to match a query image, then the color matching accuracy is improved, but large computational resources are required and significant processing delays occur
Solution Approach 1:
Instead of recoloring the entire image (which is computationally expensive), the patent extracts color histograms from both the query image and candidate images, and compares these extracted features. This extraction approach maintains color matching accuracy while avoiding the heavy computational burden of actual image recoloring.
Solution Approach 2:
The patent replaces the mechanical process of actual image recoloring with a computational comparison of color histogram features. By substituting the physical transformation of image pixels with a mathematical comparison of color distribution vectors, the system achieves accurate color matching without the processing delays associated with real-time image manipulation.
3Ease of operation
If the image search system uses user-specified colors (RGB values or Hex values) to search for images, then the search process is simple and fast, but the results are unreliable due to variation in color perception between observers
Solution Approach 1:
The patent transforms color representation from device-dependent parameters (RGB values, Hex codes) to a perceptually uniform color space (such as CIELAB). This parameter transformation ensures that equal distances in the color space correspond to equal perceived color differences, making the search results reliable across different observers and devices while maintaining the simplicity of the search process.
Data Source
AI summary
In implementations of multi-resolution color-based image search, an image search system determines a color vector for a query image based on a color histogram of the query image by concatenating two color histograms having different resolutions. The image search system can compute distance measures between the color vector of the query image and color vectors of candidate images. The image search system can select one or more of the candidate images to return based on the distance measures utilizing the distance measures as indication of color similarity of the candidate images to the query image.