Blurriness Detection via RMSE Slope for Depth Sensing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional photographic devices lack depth detection capabilities, leading to inaccurate object detection due to sensitivity to light source changes, as they rely on estimating depth from blurriness values which are unstable under varying light conditions.
Innovation Solution
A method and device that generate image blocks, filter them using gradually enhancing parameters to compute root mean square error (RMSE) values, and calculate a slope to determine blurriness, allowing for accurate detection of sharp objects by comparing these values with a threshold, independent of light source changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional blurriness detection methods are used, then object detection can be performed, but the detection results are sensitive to light source changes and lack stability
Solution Approach 1:
The patent transforms the blurriness detection from direct pixel intensity comparison to slope calculation based on RMSE values. By computing the slope of the approximate line formed by RMSE values at different filtering levels, the method changes the measurement parameter from absolute intensity difference to rate of change, which is inherently more robust to lighting variations.
Solution Approach 2:
The patent introduces multiple filtering parameters and RMSE computation as intermediary steps between the original image and the final blurriness measurement. These intermediaries (filtering operations with progressively stronger parameters, RMSE calculations) serve as a buffer that isolates the final detection from direct light source variations, allowing stable slope-based blurriness assessment.
2Loss of information
If additional depth detecting devices are added, then depth information can be obtained, but device complexity increases
Solution Approach 1:
The patent enables the imaging system to extract depth information from its own captured image data through computational processing. By analyzing blurriness slopes across different regions of the captured image, the system self-generates depth information without requiring external depth sensing hardware, thus avoiding increased device complexity while recovering depth information.
Solution Approach 2:
The patent replaces physical depth detection mechanisms (such as additional sensors or structured light projects) with computational image processing methods. The mechanical/optical depth detection system is substituted with an algorithmic approach that derives depth from blurriness analysis, simplifying the overall system architecture.
3Ease of operation
If edge-based blurriness computation is used, then object detection can be performed, but the method is sensitive to light source changes causing varying blurriness values
Solution Approach 1:
The patent segments the image into multiple filtering levels with progressively stronger parameters, computing RMSE at each level. This segmentation creates a multi-scale analysis framework where the slope calculation integrates information across scales, making the detection less sensitive to local light variations and more robust overall.
Data Source
AI summary
A method for detecting objects is provided. The method comprises the steps outlined below. An image having pixels is acquired. Image blocks each corresponding to one of the pixels are generated. A specific image block is filtered using N filtering parameters that gradually enhance the blurriness of the specific image block to generate N filtering results. N RMSE values are computed, in which the M-th RMSE value is computed according to the M-th and the (M−1)-th filtering results. A slope of an approximate line is computed according to the RMSE values as the blurriness value of the specific image block. The above steps are repeated to generate the blurriness values of all the pixels. The blurriness value is compared to a threshold value to detect sharp pixels which are parts of a sharp object and further detect an in-focus object.