This technology is a method and system that performs median filter denoising by calculating the noise level in the HH subband through wavelet transformation of an input image and determining the median filter block size k based on that value.
While 2D median filters have been widely used to remove noise while preserving details, they suffer from reduced effectiveness when fixed block sizes are not matched to the varying noise levels of different images. Furthermore, there has been no standard for determining an appropriate block size for images where the noise level is unknown.
This technology calculates the noise standard deviation using the median of the wavelet transform's high-high (HH) component and links it to the block size via a formula. It applies a larger block for stronger filtering when noise is high, and a smaller block to preserve details when noise is low. It automatically finds the optimal size even for images without prior noise information. Applicable to photo editing software and digital camera post-processing engines, it ensures consistent quality for photos taken under varying conditions without manual user adjustment.
N/A
