METHOD FOR TUNING IMAGE SIGNAL PROCESSOR AND ELECTRONIC DEVICE

Smart overview of the Invention

The patent application outlines a method for tuning an image signal processor (ISP) using an electronic device. This method involves obtaining an image evaluation model that processes images and provides evaluations. Key features of the images are determined based on these evaluations, leading to the creation of a parameter list. A tuning model, employing reinforcement learning, is trained using this parameter list and the image evaluation model to output parameter adjustment sets for image processing.

Background

Camera systems aim to deliver high-quality images by processing raw data from image sensors using ISPs. These processors consist of numerous image processing blocks with thousands of parameters, making manual tuning by engineers time-consuming and costly. The invention seeks to enhance image quality while reducing the time and complexity associated with tuning ISPs.

Summary of Invention

The invention offers a method for tuning ISPs through an electronic device, improving performance and reducing processing time. It involves training a tuning model using an image evaluation model and a parameter list derived from key image features. This model outputs parameter adjustment sets to optimize image processing, leveraging reinforcement learning techniques.

System Components

The system comprises a tuning system with a first electronic device generating tuning parameters and a second device applying these to enhance image processing. The first device creates preset tuning parameters using models like artificial neural networks or decision trees. These presets are stored in a database and accessed by the second device, which applies them to input images to produce optimized outputs.

Application and Benefits

The described method allows for objective and accurate ISP tuning, overcoming the limitations of human evaluation. By generating optimal parameter sets and output images, the system enhances ISP performance and aligns image processing with user preferences. This approach is applicable to various devices, including smartphones, tablets, and wearable technology, enhancing image quality across a wide range of platforms.