Frequency-Domain Image Analysis Model for Fast Focus Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional surveillance cameras face challenges in efficiently and accurately determining the focus state of detection images due to the large amount of spatial domain information requiring significant memory and computation resources, leading to lengthy processing times.
Innovation Solution
The method involves dividing detection images into sub-images and transforming them into a frequency domain using discrete cosine transformation, applying masks and filters through a fully connected multilayer perceptron network to generate an analysis model output layer, and adjusting parameters based on comparison results to optimize the prediction process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If spatial domain information is used to determine focus state, then measurement precision is improved, but device complexity and computation time increase significantly
Solution Approach 1:
The patent transforms image data from spatial domain to frequency domain using Discrete Cosine Transformation (DCT). This dimensionality change allows focus determination to be performed on frequency coefficients rather than spatial pixels, significantly reducing computational complexity while maintaining measurement precision. The frequency domain representation compactly encodes image content, enabling efficient focus metric calculation.
Solution Approach 2:
The patent changes the representation parameters of image data by applying DCT transformation. Instead of operating on spatial domain pixel values, the system operates on frequency domain coefficients (DC and AC coefficients). This parameter transformation reduces the amount of data to be processed and enables simpler computational operations for focus determination.
2Measurement precision
If spatial domain information is used to determine focus state, then measurement precision is improved, but loss of time increases due to lengthy computation
Solution Approach 1:
By transforming to frequency domain, the patent enables faster computation of focus metrics. The DCT-transformed coefficients can be processed more efficiently than spatial pixels, reducing the time required for focus determination while preserving measurement accuracy through the energy-compacting property of DCT.
Solution Approach 2:
The patent extracts only the essential frequency components (DC and selected AC coefficients) from the full image data. This extraction of critical information eliminates the need to process all spatial pixels, significantly reducing computation time while maintaining sufficient accuracy for focus determination.
3Measurement precision
If spatial domain information is used, then measurement precision is improved, but use of energy increases due to large memory requirements
Solution Approach 1:
The patent changes the data representation from spatial pixels to frequency coefficients through DCT transformation. This parameter change results in a more compact representation that requires less memory storage, thereby reducing the energy consumption associated with memory operations while maintaining the ability to accurately determine focus state.
Solution Approach 2:
The patent extracts and retains only the most significant frequency coefficients (DC and selected AC coefficients) needed for focus determination. This selective extraction reduces the volume of data that must be stored in memory, decreasing both memory requirements and the associated energy consumption.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables rapid and accurate determination of image focus state, reducing computational complexity and maintaining high efficiency and accuracy by leveraging frequency domain analysis and machine learning.
Implementation Method 1
transforming them into a frequency domain using discrete cosine transformation
Data Source
AI summary
An image analysis model establishment method is applied to an image analysis apparatus including an imager receiver and an operation processor. The image analysis model establishment method divides a detection image acquired by the imager receiver into a plurality of sub-images and transforms the sub-images from a spatial domain to a frequency domain to generate a plurality of pre-processing frequency domain data, generating an analysis model output layer by an inner product of the plurality of pre-processing frequency domain data transformed via a plurality of masks and several filters through a fully connected multilayer perceptron network, acquiring a predicted result of the detection image in accordance with a category determination result of the analysis model output layer, and comparing the predicted result with a target label to adjust parameters in each transformation phase applied for the plurality of pre-processing frequency domain data in accordance with a comparison result.