The denoising of signals often involves the reduction of pulse-type noise generated by electronic CCD / CMOS sensors or the acquisition electronics associated with them. The size of the processed image is limited because current solutions (median / convolution filters) have limitations in terms of processing speed and sizes of noise filters.
The invention provides techniques for implementing and optimizing median / convolution / iterative filters for GPUs. This results in advantageous performance in terms of processing speed, the size of the filters allowed, the quality of the images (for the medians). In addition, the method is applicable to 1D (audio) or 2D (images) signals, and possibly 3D.
MARKETS AND APPLICATIONS
Validation of the technique in an operational environment (TRL 6)
Institut FEMTO-Sciences & Technologies
Software deposit in progress
License or co-development
Business Development Manager
+33 (0)6 26 61 89 06