DECONVOLUTION APPARATUS AND METHOD USING A LOCAL SIGNAL-TO-NOISE RATIO

• Main Authors: ZIESCHE, Florian, REYMANN, Jürgen, WALTER, Kai
• Format: Patent
• Language: English; French; German
• Published: 10.08.2022
• Subjects: CALCULATING; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The invention relates to a method and an apparatus, such as an image processor (1), for deconvolving ann-dimensional digital input imagesI(xi), whereinnis at least two. The digital input image comprises a plurality of input voxels (xi). The deconvolution according to the invention uses a noise componentβ(xi) depending on a local signal-to-noise ratio (SNR(xi)) which is computed from a noise levelN(xi) and a signal levelS(xi). Both the signal level and the noise level are computed by using only voxels of the input image that are located in an input regionR(xi) surrounding the input voxel, at which the current signal-to-noise ratio is determined. The noise component is limited to a maximum noise value (βmax) for a signal-to-noise ratio below a lower SNR threshold (SNRmin) and to a minimum noise value (βmin) for a signal-to-noise ratio above an upper SNR threshold (SNRmax). The deconvolution may be a Wiener or a maximum a posteriori deconvolution, such as a Lucy-Richardson deconvolution. The noise level is preferably determined as the variance of a derivative value of the input image, so that a summed-area table may be used. The apparatus and method of the invention are in particular suited for images from a microscope or endoscope recording in particular fluorescence images.