Wavelet power spectrum-based autofocusing algorithm for time delayed and integration charge coupled device space camera

Abstract

A novel autofocusing algorithm using the directional wavelet power spectrum is proposed for time delayed and integration charge coupled device (TDI CCD) space cameras, which overcomes the difficulty of focus measure for the real-time change of imaging scenes. Using the multiresolution and band-pass characteristics of wavelet transform to improve the power spectrum based on fast Fourier transform (FFT), the wavelet power spectrum is less sensitive to the variance of scenes. Moreover, the new focus measure can effectively eliminate the impact of image motion mismatching by the directional selection. We test the proposed method's performance on synthetic images as well as a real ground experiment for one TDI CCD prototype camera, and compare it with the focus measure based on the existing FFT spectrum. The simulation results show that the new focus measure can effectively express the defocused states for the real remote sensing images. The error ratio is only 0.112, while the prevalent algorithm based on the FFT spectrum is as high as 0.4. Compared with the FFT-based method, the proposed algorithm performs at a high reliability in the real imaging experiments, where it reduces the instability from 0.600 to 0.161. Two experimental results demonstrate that the proposed algorithm has the characteristics of good monotonicity, high sensitivity, and accuracy. The new algorithm can satisfy the autofocusing requirements for TDI CCD space cameras.

Autofocus method for scanning remote sensing cameras

“…Nevertheless, the effectiveness of these methods depends on the consistency of the scenes acquired during the autofocus process and may lead to incorrect instructions, since they are not on the basis of the same scene. Further, as mentioned in [11], the error ratio of these methods can be as high as 11.2% for a wavelet-power-spectrum-based focus measure and 40% for an FFT-power-spectrum-based focus measure. These are far from common acceptance for autofocus application of remote sensing cameras.…”

“…Further, since the DWT-based method [11] has shown the best performance so far, a comparison is carried out between our method and the DWT-based method. In the following experiments, the same focus measure (proposed in [11]) is used in the two methods.…”

“…This is essentially because the images are not based on the same scene in the DWT-based method. As mentioned in [11], the most arbitrary scenes have the same FFT power spectra and wavelet power spectra, but they are not exactly the same. Figures 5 to 7 indicate that, in each focusing step, the wavelet power spectra values of the eight images obtained at the same focal position are different.…”

“…Accordingly, this method is not applicable for high-speed, fine-resolution remote sensing cameras and has considerable difficulties even for lower resolution cameras. Some researchers have turned to focus measures that are not dramatically sensitive to varying scenes, such as a power-spectrum-based focus measure [10] and wavelet-powerspectrum-based focus measure [11]. Nevertheless, the effectiveness of these methods depends on the consistency of the scenes acquired during the autofocus process and may lead to incorrect instructions, since they are not on the basis of the same scene.…”

Analysis of image motion on autofocus precision for aerial cameras

“…[1][2][3][4][5][6][7][8][9][10] The image-based passive autofocus algorithms are widely studied for their high accuracy and no requirement for extra equipment. [1][2][3][4][5][6][7][8][9][10] The image-based passive autofocus algorithms are widely studied for their high accuracy and no requirement for extra equipment.…”