Signal processing device, signal processing method, recording medium
The signal processing device enhances image recognition systems by converting images from portable cameras to match surveillance camera quality, addressing differences in image characteristics and improving AI model inference accuracy through RAW image simulation and signal processing.
Patent Information
- Authority / Receiving Office
- JP Β· JP
- Patent Type
- Applications
- Current Assignee / Owner
- SONY SEMICON SOLUTIONS CORP
- Filing Date
- 2024-11-22
- Publication Date
- 2026-06-03
AI Technical Summary
Existing image recognition systems using AI models on surveillance cameras face challenges in converting images captured by non-surveillance cameras into surveillance camera quality, resulting in differences in image quality, which can degrade inference performance due to differences in image characteristics between input and training images, and existing technologies have not addressed the issue of converting images between cameras, resulting in differences in image characteristics between training and training images, and existing technologies have not effectively simulated images to match the quality of surveillance cameras.
A signal processing device that includes a RAW image simulation unit and a signal processing simulation unit to convert images captured by a portable camera into images with the same quality as if they were captured by a surveillance camera, by applying imaging characteristics and inverse signal processing to simulate a RAW image with the image quality of surveillance cameras.
The solution effectively addresses the challenge of enhancing image recognition systems by enabling the conversion of captured images, enhancing the accuracy of image recognition systems, and enhancing the efficacy of image recognition systems, and enhancing the efficacy of image recognition processes, and improving the inference accuracy of AI models through simulated images.
Smart Images
Figure 2026090863000001_ABST