A mixed statistical model to evaluate a common image sensor defect.
One of the most common defects in image sensor devices is spatial blotches. The appearance of blotches in image sensors is a regular occurrence and may be generated by internal moving parts or may be moved by air currents within the camera.
Composed of two main statistical methods, the first module employs an inferential method, applying a spatial segmentation of the current frame to obtain candidate spot regions by testing the null hypothesis of each sample. Subsequently, a regression method is added and is applied only on the candidate spots in order to separate the bad pixels that generate the dust spots. Finally, the proposed mixed model is applied under real conditions in order to evaluate its performance in detection stages using a real image capture.
Watch a video demonstrating the process from the Test Vision Symposium.
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
Leave a Reply