Author Affiliations
1 School of Energy and Power Engineering, Xi'an Jiaotong University, Xi'an, Shaanxi 710049, China2 The State Key Laboratory of Laser Interaction with Matter, Northwest Institute of Nuclear Technology, Xi'an, Shaanxi 710024, Chinashow less
Fig. 1. HTV experimental images. (a) OHB fluorescence interference; (b) flow field noise; (c) system noise
Fig. 2. Background noise. (a) Schematic of experimental setup; (b) experimental results
Fig. 3. Contrast of background noise in and out of PLIF display sheet
Fig. 4. Simulation images with background noise. (a) Simulation model with odd inflection points; (b) simulation model with even inflection points
Fig. 5. Background noise histogram image
Fig. 6. Standard deviation versus gain
Fig. 7. Contrast of simulation results and experimental results. (a) Simulation results; (b) experimental results
Fig. 8. Contrast of de-noising results of different parameters of simulation models. (a) Contrast of de-noising results with different window numbers with odd number of inflection points; (b) contrast of de-noising results with different window numbers with even number of inflection points; (c) evaluation of de-noising results of different signal models; (d) evaluation of de-noising results with different window sizes
Fig. 9. De-noising results of noise model. (a) Noise model; (b) ROI; (c) Hough detection; (d) partition filtering
Fig. 10. De-noising analysis of wavelet transform. (a)Model; (b)noised model; (c) RPSNR versus wavelet threshold coefficient; (d) RPSNR versus wavelet decomposition layer; (e) RPSNR versus wavelet iterated time; (f) RPSNR 、RSNR of de-noised signal versus RSNR of noised signal; (g) wavelet transform processing
Fig. 11. Flow diagram of image preprocessing
Fig. 12. Contrast of simulation de-noising results. (a) Noised model; (b) Gaussian smoothing; (c) median filtering; (d) Laplace sharpening; (e) spatial filtering; (f) proposed method
Fig. 13. Experimental locale photo
Fig. 14. De-noising results of experimental data. (a) Experimental image; (b) proposed method result; (c) Gaussian smoothing result; (d) median filtering result; (e) Laplace sharpening result
Noise type | OHB interference | Flow field noise | System noise |
---|
Morphological characteristic | Concentrated | Discrete | Discrete | Power | Highest, concentrated | Higher, dispersed | High, dispersed | Position space | In PLIF display sheet,uncovered the signal | Distributed throughout theimage, especially in thePLIF display sheet | Distributed throughoutthe image | Existing reason | Occured in areas of combustionor chemical reactions | Caused by scattering of the flowfield wall or other particles | Existed in allexperimental images |
|
Table 1. Characteristic analysis of background noise
Method | Spatial filter based on Hough transform | Wavelet transform |
---|
Feature | Less calculation, can segment differentmorphology distribution images | Can remove Gaussian noise of image | Applied object | Stray light in flow field and uncoveredsignal combustion interference | Sensor noise including physical noise,random noise and so on | Disadvantage | Not suitable for suppressing background ofsignal covered by interference | Not suitable for interference and noisefrom combustion flow field | Solution | When removing background noise,combines with wavelet transform | Optimizes spatial filter |
|
Table 2. Background noise processing methods
Method | RPSNR/dB | RSNR/dB | RMSE/dB | RPDE/% |
---|
Original image | 11.9121 | 0.6626 | 4.1867×103 | 27.12 | Gaussian smoothing | 14.0240 | 1.2696 | 2.5744×103 | 26.45 | Median filtering | 13.7175 | 1.2129 | 2.7627×103 | 26.84 | Laplace sharpening | 13.0269 | 1.4407 | 3.2388×103 | 26.97 | Spatial filtering | 25.9605 | 9.4139 | 164.8278 | 0.846 | Spatial filtering & wavelet function | 28.7014 | 14.5683 | 87.6880 | 0.711 |
|
Table 3. Contrast of de-noising results