Image Encoding Region Estimation for Surveillance Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding techniques for surveillance cameras face challenges in maintaining high image quality for specific regions, such as human faces, due to delays in detection processes and incorrect region prediction, leading to degraded image quality and increased code size.
Innovation Solution
An image processing apparatus that includes a detection unit for identifying specific regions, an estimation unit for predicting their location, a calculation unit for measuring shifts, and a setting unit that adjusts encoding parameters to enhance image quality for predicted regions within a predetermined range, while correcting errors outside this range.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the detection process is executed repeatedly to improve detection accuracy, then the reliability of specific region detection is improved, but the loss of time increases due to delay in encoding process
Solution Approach 1:
The estimation unit performs preliminary estimation of the specific region before the detection process completes. By predicting the region location in advance based on historical data and motion models, the system can prepare encoding parameters ahead of time, reducing the delay caused by waiting for detection completion while maintaining high detection reliability through repeated detection processes.
2Manufacturing precision
If the encoding parameter is changed for the delayed specific region, then the image quality of the specific region is improved, but the code amount of the whole scene increases
Solution Approach 1:
The system applies different encoding qualities to different regions. The estimation unit identifies the specific region and applies high-quality encoding parameters only to that local area, while maintaining standard encoding parameters for the rest of the scene. This localized approach improves the image quality of the specific region (such as human face) without unnecessarily increasing the code amount for the entire scene.
3Productivity
If the specific region is predicted based on delayed detection results, then the encoding process can proceed without waiting for detection completion, but the prediction accuracy decreases leading to wrong region identification
Solution Approach 1:
The system uses feedback from repeated detection processes to refine the prediction. The estimation unit continuously compares predicted regions with actual detection results from multiple detection cycles, adjusting the prediction based on the feedback. This feedback mechanism allows the system to maintain high prediction accuracy even while proceeding with encoding before detection fully completes, preventing wrong region identification.
Data Source
AI summary
A specific region in each frame image input after a detection process has been completed a predetermined number of times or more is estimated from a specific region detected from a past frame before the frame. The shift between the specific region detected from a first frame image input after the detection process has been completed the predetermined number of times or more and the specific region estimated for the first frame image is obtained. When the shift falls within a predetermined range, an encoding parameter to encode the specific region estimated for a second frame image input at a point the detection process for the first frame image has been completed with a higher image quality than that of regions other than the specific region is set.