This is an old revision of the document!


OptimizeReferences

This logic aligns the class averages so that they are in the same direction. After classification step different classes might contain the same view, but rotated in the image plane, so alignment is needed for the same views to look identical.

The quality of the output heavily depends on the quality of the first class average chosen - preferably, it should show enough details of the molecule or complex and be surrounded by a black rim. The optimal values suggested for radius parameter are around 0.8-0.9.

Parameters Description
Correlation Function The correlation function that is optimized during the image alignment
Fraction The maximal fraction of the image that will be shifted relative to reference image during the alignment
Interpolation Defines the interpolation function used to get the value between two neighbouring pixels
Radius The value of Circular Mask Fraction (this fraction of image dimension will be used as diameter of the circle cut from the image for cf calculations)
Sampling The maximal fraction of the image shifted during the alignment relative to the reference image
Input Description
Input A set of class averages
Output Description
Output A set of aligned class averages
Written Header Values Description
resolutionAt0.143 Estimated resolution for FSC cut off at 0.143
resolutionAt0.3 Estimated resolution for FSC cut off at 0.3
resolutionAt0.5 Estimated resolution for FSC cut off at 0.5
resolutionAt0.9 Estimated resolution for FSC cut off at 0.9
resolutionAt3Sigma Estimated resolution for FSC cut off at 3 Sigma
resolution Estimated resolution at the user chosen cut off value
maxResolution Maximum resolution in the local map
meanResolution Mean resolution in the local map
minResolution Minimum resolution in the local map
resolutionVariance Variance of the resolution in the local map
abscissaIsLogarithmical , abscissaMin, abscissaMax For internal 1D viewer use only, to be hidden

FIXME

The logic takes a stack of class average images as input. It applies the circular mask of radius changed by parameter 'Radius' to every image, so that only the defined circle and not the surrounding noise will be used for correlation calculations. Then the logic aligns every image to the first one by shifting them in polar coordinates (that corresponds to rotation of image in Cartesian coordinates) and maximizing the correlation function.