Scene matching NCC value improvement based on contrast matching
| Autor: | Sajad Poursajadi, Saeed Karimifar, Ali Pourmohammad |
|---|---|
| Rok vydání: | 2013 |
| Předmět: |
Matching (statistics)
Mean squared error business.industry media_common.quotation_subject ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Histogram matching Pattern recognition Edge detection Computer Science::Computer Vision and Pattern Recognition Histogram Contrast (vision) Adaptive histogram equalization Computer vision Artificial intelligence business Image histogram media_common Mathematics |
| Zdroj: | 2013 8th Iranian Conference on Machine Vision and Image Processing (MVIP). |
| Popis: | Geometrical and radiometrical corrections are important for scene matching applications. We suppose the applications that there are no geometrical errors based on using 3D-Inertial sensors for geometrical corrections. In these cases, Normalized Cross-Correlation (NCC) is commonly used method for scene matching. The problem of matching a pattern image (mask) to an image in these cases needs to correction of radiometrical errors as illumination (contrast) variations. In this paper we show that correlation between a mask and a histogram matched image instead of using that raw version, improves the correlation value. First we match histogram function of the image to histogram function of the mask in order to have two closed contrast images, and then correlate those together using NCC and root mean square error (RMSE) methods. Simulation results confirm that according to using NCC and RMSE simultaneously, not only this method is a fast and real time method, but also according to matching histogram function of the received image to histogram function of the mask, it improves the correlation value. Also we show that using the edge detected version of the mask and histogram matched image, lead us to have the best results. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|