Adam Lüning och Erik Chen presenterar sitt examensarbete "Design of Image Quality Test Bench for Infrared and Low-light Cameras"

Titel: Design of Image Quality Test Bench for Infrared and Low-light Cameras

Abstract: 

Objective Image Quality Assessment (IQA) is crucial for evaluating and ensur- ing the fidelity of images. Traditional human-based evaluations, while reliable, are impractical due to their cost and slow pace, coupled with subjective incon- sistencies. This thesis aims to address these issues by developing a test and quality bench for Bertin Exensor’s dual-head camera, integrating both infrared and low-light cameras. The goal is to create an efficient, standardized, and ob- jective quality testing system to enhance consistency and reliability in evaluating the camera heads that arrive at Bertin Exensor’s office.

The project is divided into two main parts: evaluation of IQA methods and design of the test bench. The chosen IQA methods were selected for their alignment with human judgment, focusing primarily on sharpness, identified as the most relevant factor for image quality. The test bench was set in a controlled environment and was designed as a user-friendly graphical interface, providing a simple pass/fail outcome based on the created IQA algorithm, ensuring ease of use for operators.

Results indicate that while assessment of sharpness in the center of the image proves reliable, evaluating overall image quality remains limited in accuracy. The project faced challenges due to the limited availability and inconsistent quality distribution of camera heads, affecting threshold optimization and fair comparison capabilities. Despite these limitations, the test bench demonstrated high accuracy in evaluating image sharpness and proved effective in minimizing misclassification risks through a majority voting approach.

The thesis concludes that the test bench offers a quick and objective evaluation of dual-head cameras, closely aligning with human perception. However, fur- ther validation on a larger number of camera heads is necessary to ensure high accuracy and robustness of the algorithms.

Examinator: Gabrielle Flood

Handledare: Marie Requirand och Magnus Oskarsson