帮忙翻译汉译英

Question

帮忙把下面一段文字翻译成英文，多谢
从均匀辐射的原始图像可以看出，长波IRFPA具有较大的低频空间噪声，主要体现为偏置噪声（图像上的固定图案）。考察传统神经网络算法的期望值计算公式，像点的期望输出为校正输出的四邻域均值。这样对于这种偏置占优的场合，估计的期望值与真实值之间的差距就会很大。因为：神经网络算法对于原始图像中的高频空间噪声很有很好的抑制作用。因为此类噪声的空间相关性比较弱，所以可以通过线性平滑环节将这些高频的噪声和读出电路的非均匀性噪声进行滤除。而对于IRFPA的非均匀性在空间频域种主要是低频噪声，这类噪声的空间相关性较强，场景中不断连续变化的图像也具有很强的空间相关性，通过线性平滑环节，就可以保留这些高相关性的场景信息。但是同时也保留下了低频噪声信息。这就造成了传统的神经网络算法无法较好的对偏置噪声进行校正。从上面的校正后后的图像也可以看出，固定的偏置成分通过较长帧的校正后，仍然存在，这点可以比较校正前后的两幅图像（图8和图9）可以看出。

Answer 1

Uniform radiation from the original image can be seen, long-wave IRFPA greater low-frequency noise in space, mainly reflected in bias noise (image on the fixed pattern). Study of traditional neural network algorithm formula for calculating the expected value, such as the desired output point for the neighborhood domain Calibration output mean. This is such a dominant bias occasion, it is estimated that the expectations and the gap between the true value would be great. Since: neural network algorithm for the original image is very good high-frequency noise Spaces inhibition. Since such noise is relatively weak spatial correlation, so can linear smoothing aspect of these high-frequency noise and ROIC nonuniformity noise filter. For IRFPA nonuniformity in space are the main types of frequency-domain low-frequency noise, this type of noise in a strong spatial correlation, the scene changes continuously consecutive image also has a strong spatial correlation, by linear smoothing aspect, on the be able to retain the relevance of these high-scenes information. But at the same time retain the information under the low-frequency noise. This has resulted in the traditional neural network algorithm can not be good for Calibration of the bias noise. After correction from the above image also can be seen that the fixed component of the bias correction through the longer frames still exist, this point can compare the two images before and after correction (Figure 8 and Figure 9) can be seen.

Answer 2

想努力，但太难

Answer 3

Uniform radiation from the original image can be seen, long-wave IRFPA greater low-frequency noise in space, mainly reflected in bias noise (image on the fixed pattern). Study of traditional neural network algorithm formula for calculating the expected value, such as the desired output point for the neighborhood domain Calibration output mean. This is such a dominant bias occasion, it is estimated that the expectations and the gap between the true value would be great. Since: neural network algorithm for the original image is very good high-frequency noise Spaces inhibition. Since such noise is relatively weak spatial correlation, so can linear smoothing aspect of these high-frequency noise and ROIC nonuniformity noise filter. For IRFPA nonuniformity in space are the main types of frequency-domain low-frequency noise, this type of noise in a strong spatial correlation, the scene changes continuously consecutive image also has a strong spatial correlation, by linear smoothing aspect, on the be able to retain the relevance of these high-scenes information. But at the same time retain the information under the low-frequency noise. This has resulted in the traditional neural network algorithm can not be good for Calibration of the bias noise. After correction from the above image also can be seen that the fixed component of the bias correction through the longer frames still exist, this point can compare the two images before and after correction (Figure 8 and Figure 9) can be seen.

Answer 4

Uniform radiation from the original image can be seen, long wave IRFPA has the big low frequency spatial noise, mainly manifests for the bias noise (on image fixed design). Inspection tradition neural network algorithm expected value formula, image point expectation output for adjustment output four neighborhood average value. Like this occupies the superior situation regarding this kind of bias, between the estimate expected value and the real value disparity can be very big. Because: The neural network algorithm has the very good inhibitory action very much regarding the primitive image's in high frequency spatial noise. Because is quite weak for this kind of noise's spatial relevance, therefore may carries on through the linear smoothing link these high frequency noises and reading circuit's heterogeneous noise the filtration. But plants regarding the IRFPA heterogeneity in the spatial frequency range is mainly the low-frequency noise, this kind of noise's spatial relevance is strong, in the scene changes continuously unceasingly the image also has the very strong spatial relevance, through the linear smoothing link, may retain these high relevant scene information. But simultaneously also the retention has gotten down the low-frequency noise information. This caused the traditional neural network algorithm to be unable to carry on the adjustment well to the bias noise. After the above adjustment the image may also see, fixed bias ingredient after long frame adjustment, still existed, around this may compare adjusts two images (Figure 8 and Figure 9) may see.

Answer 5

As the original image shows the even radiation, IRFPA with longer wave length have higher low frequency space noise, which mainly expressed as biased noise (the fixed pattern in the image). By examining the formula for calculating the expected values of traditional neuro-network,

这是关於 neuro-network image processing 的文章，如果楼主能把一些华语专业词的英文写出来，我可以试试。

像点、四邻域、线性平滑环节、场景信息

Answer 6

May see from homogeneous radiation's primitive image, long wave IRFPA has the big low frequency spatial noise, mainly manifests for the bias noise (on image fixed design). Inspection tradition neural network algorithm expected value formula, image point expectation output for adjustment output four neighborhood average value. Like this occupies the superior situation regarding this kind of bias, between the estimate expected value and the real value disparity can be very big. Because: The neural network algorithm has the very good inhibitory action very much regarding the primitive image's in high frequency spatial noise. Because is quite weak for this kind of noise's spatial relevance, therefore may carries on through the linear smoothing link these high frequency noises and reading circuit's heterogeneous noise the filtration. But plants regarding the IRFPA heterogeneity in the spatial frequency range is mainly the low-frequency noise, this kind of noise's spatial relevance is strong, in the scene changes continuously unceasingly the image also has the very strong spatial relevance, through the linear smoothing link, may retain these high relevant scene information. But simultaneously also the retention has gotten down the low-frequency noise information. This caused the traditional neural network algorithm to be unable to carry on the adjustment well to the bias noise. After the above adjustment the image may also see, fixed bias ingredient after long frame adjustment, still existed, around this may compare adjusts two images (Figure 8 and Figure 9) may see.