Skip to the navigation links
Last modified: 3 April 2024


Bugs: aconvolve

Table of Contents



The 2D raised cosine and hanning convolution kernels are not correct.

The 1D versions are correct.

Byte type images (unsigned 8-bit integer) produce incorrect results when using method=fft

When using method=fft, byte type images are incorrectly cast to a singed data-type, meaning values above 128 become negative.


Users can either user method=slide or change the data-type by specifying it as part of the virtual file syntax, eg

aconvovle "infile.img[opt type=i2]" ...