Digital image correlation with self-adaptive scheme for interpolation bias reduction
| Autor: | Peihan Tu |
|---|---|
| Rok vydání: | 2017 |
| Předmět: |
Demosaicing
Computer science Image quality Applied Mathematics ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Bilinear interpolation Stairstep interpolation 02 engineering and technology 021001 nanoscience & nanotechnology 01 natural sciences Multivariate interpolation 010309 optics 0103 physical sciences Image noise 0210 nano-technology Instrumentation Engineering (miscellaneous) Algorithm Image gradient Interpolation |
| Zdroj: | Measurement Science and Technology. 28:075008 |
| ISSN: | 1361-6501 0957-0233 |
| DOI: | 10.1088/1361-6501/aa70f8 |
| Popis: | In digital image correlation (DIC), the systematic error caused by intensity interpolation at sub-pixel positions, namely the overall interpolation bias, includes both interpolation bias and noise-induced bias. The overall interpolation bias is especially significant when the noise level is high or the image contrast is low. There is a pressing need to reduce the overall interpolation bias to improve the accuracy of DIC. However, existing approaches such as using a low-pass filter or a high-order interpolation require manually selected algorithm parameters, and cannot reduce the bias automatically. It is known that the overall interpolation bias is highly correlated with image gradient (and thus the contrast of the speckle image). This provides an opportunity to reduce the bias simply by adjusting the gradients. Inspired by the image enhancement technique which is used to alter image gradients (thus image contrast) by nonlinearly transforming its intensities (RGB, gray-value, etc), a DIC algorithm called the gray-level adaptive DIC (GA-DIC), based on a new correlation criterion with an additional adjustable parameter which controls the gradients, is proposed to reduce the overall interpolation bias. Both numerical and real experiments are applied to verify the feasibility and effectiveness of the GA-DIC. The results show that the proposed algorithm can reduce the overall interpolation bias without empirically selecting algorithm hyperparameters. Its effect is more significant in cases with higher image noise and poorer image quality. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|