裁剪文件

Question

考虑以下代码（下载test.fits）：

from astropy.io import fits
from photutils.utils import cutout_footprint

# Read fits file.
hdulist = fits.open('test.fits')
hdu_data = hdulist[0].data
hdulist.close()

# Some center and box to crop
xc, yc, xbox, ybox = 267., 280., 50., 100.
# Crop image.
hdu_crop = cutout_footprint(hdu_data, (xc, yc), (ybox, xbox))[0]
# Add comment to header
prihdr = hdulist[0].header
prihdr['COMMENT'] = "= Cropped fits file")
# Write cropped frame to new fits file.
fits.writeto('crop.fits', hdu_crop, prihdr)

原始（左）和裁剪（右）图像如下所示：

框架中心的星星的（ra，dec）equatorial coordinates是：

Original frame: 12:10:32  +39:24:17
Cropped frame:  12:12:07  +39:06:50

为什么裁剪框架中的坐标不同？

这是使用两种不同方法解决此问题的两种方法。

from astropy.io import fits
from photutils.utils import cutout_footprint
from astropy.wcs import WCS
from astropy.nddata.utils import Cutout2D
import datetime

# Read fits file.
hdulist = fits.open('test.fits')
hdu = hdulist[0].data
# Header
hdr = hdulist[0].header
hdulist.close()

# Some center and box to crop
xc, yc, xbox, ybox = 267., 280., 50., 100.

# First method using cutout_footprint
# Crop image.
hdu_crop = cutout_footprint(hdu, (xc, yc), (ybox, xbox))[0]
# Read original WCS
wcs = WCS(hdr)
# Cropped WCS
wcs_cropped = wcs[yc - ybox // 2:yc + ybox // 2, xc - xbox // 2:xc + xbox // 2]
# Update WCS in header
hdr.update(wcs_cropped.to_header())
# Add comment to header
hdr['COMMENT'] = "= Cropped fits file ({}).".format(datetime.date.today())
# Write cropped frame to new fits file.
fits.writeto('crop.fits', hdu_crop, hdr)

# Second method using Cutout2D
# Crop image
hdu_crop = Cutout2D(hdu, (xc, yc), (xbox, ybox), wcs=WCS(hdr))
# Cropped WCS
wcs_cropped = hdu_crop.wcs
# Update WCS in header
hdr.update(wcs_cropped.to_header())
# Add comment to header
hdr['COMMENT'] = "= Cropped fits file ({}).".format(datetime.date.today())
# Write cropped frame to new fits file.
fits.writeto('crop.fits', hdu_crop.data, hdr)

Answer 1

photutils.utils.cutout_footprint仅剪切像素，它不会更新用于在像素和世界坐标之间转换的WCS。

改为使用astropy.nddata.utils.Cutout2D。

Answer 2

The coordinates changed because you sliced the image but you did not alter the WCS informations (especially the reference pixel values).

One way would be to use astropy.WCS:

from astropy.wcs import WCS
wcs = WCS(hdulist[0].header)
wcs_cropped = wcs[280-50 : 280+50 , 267-25 : 267+25]

and then copy this updated wcs to your header:

prihdr.update(wcs_cropped.to_header())

before saving the file.

I'm not sure about what cutout_footprint does so maybe you need to change the slice indices when creating wcs_cropped.

There is an convenience functionality in astropy.nddata named Cutout2D which updates the WCS by default.

Answer 3

另一个答案是因为它需要一个额外的包：ccdproc，尤其是基于astropy.nddata.NDData的课程CCDData：

它简化了阅读文件：

from ccdproc import CCDData
ccd = CCDData.read(filename, unit='erg/cm**2/s/AA')

必须指定单位，因为标题单元与astropy.units模块不兼容。

关于CCDData的重要一点是，您（大部分）不需要照顾units，wcs，header和masks通过自己，它们被保存为属性，最基本的操作相应地保留（和更新）它们。例如切片：

xc, yc, xbox, ybox = 267., 280., 50., 100.
ccd_cropped = ccd[int(yc - ybox // 2) : int(yc + ybox // 2), 
                  int(xc - xbox // 2) : int(xc + xbox // 2)]

此切片ccd_cropped已更新WCS信息，因此您只需继续处理它（如再次保存）：

ccd_cropped.write('cropped.fits')

它应该有正确更新的坐标。切片部分实际上也可以使用astropy.nddata.NDDataRef，只有read和write部分在ccdproc.CCDData中明确实现