Changes to DIPimage
- A new option makes the functions ft and ift use the MATLAB built-in FFTW library, rather than DIPlib. This is much faster in most cases. Write dipsetpref('FFTtype','fftw').
- readtimeseries now reads series much, much faster. Can now also read files with names composed of only digits.
- readim now reads multiple images from one .lif file.
- noise: Fixed issue with 'uniform' noise and default parameters. Added blue noise.
- It is now finally possible to combine the 'tight' option and a size parameter in diptruesize.
- resample, rotation and rotation3d now can use Lanczos interpolation. 'lanczos3' is now the default interpolation method in these functions.
- rotation and rotation3d now cast the input image to a float type for all interpolation methods except 'nn', 'zoh' and 'linear'. This is to avoid clipping behaviour that was usually reported as a bug. Binary images can now be rotated with the 'nn' or 'zoh' methods.
- It is now possible to use multiple thresholds with the 'fixed' method in threshold.
- Converting a scalar value to a dip_image object no longer creates an sfloat image, the data type is always preserved.
- dip_add, dip_div, dip_mul and dip_sub have been removed in favour of the new function dip_arith, which directly corresponds with a DIPlib function. An optional input argument allows to choose the data type of the output.
- Fixed bug in convolve with 1D image and filter.
- Function psnr was computing STD instead of MSE.
- img+ gave a DIPlib error.
- eig gave errors for 3D tensors.
Changes to DIPlib
(These might or might not propagate to DIPimage)
- Added nearest neighbour interpolation (zero order hold results in half a pixel shift!) and Lanczos interpolation.
- Skews and rotations with cubic spline convolution are much faster now.
- dip_Reconstruction() now allows binary input images.
- Fixed bug that caused SegV when creating a rectangular pixel table with non-integer size.
- Arithmetic with 0D images could give a SegV.
Changes to dipIO
(This propagates to DIPimage)
- dipio_ImageReadColourSeries() reads in a series of 2D images as a single 3D image.