Excerpts from the article "Comparing linear verses non linear filters in image processing"
Embedded Computing Design,
May 2012, pp. 8-10
Real-time or embedded image processing was limited by the cost per constraints created by the silicon substrate. With the growth of sub 90 nano meter technology changed the entire scenario.
Filtering is used to do interpolation, noise reduction and resampling functions. There are two types of filtering namely linear and nonlinear. As we know Image is a two dimensional signal. Filtering theory developed for one dimensional (1D) was extented to two dimentsional field. First there was analog 1D filters then came discrete and atlast digital 1D filters. A linear system should satisfy superposition and shift invariance property. They should be causal and stable to be realized (For more information read any book on Signals and systems OR Digital Signal processing).
Non-linear filters do not satisfy superposition and shift invariance. But they should be causal and stable to be realized. Median filters and rank-order filters are very good examples for non-linear filters.
Comment: Very nice layout, read from tablet, Thanks Mr. Joshua
http:\\embedded-computing.com/
Embedded Computing Design,
May 2012, pp. 8-10
Real-time or embedded image processing was limited by the cost per constraints created by the silicon substrate. With the growth of sub 90 nano meter technology changed the entire scenario.
Filtering is used to do interpolation, noise reduction and resampling functions. There are two types of filtering namely linear and nonlinear. As we know Image is a two dimensional signal. Filtering theory developed for one dimensional (1D) was extented to two dimentsional field. First there was analog 1D filters then came discrete and atlast digital 1D filters. A linear system should satisfy superposition and shift invariance property. They should be causal and stable to be realized (For more information read any book on Signals and systems OR Digital Signal processing).
Non-linear filters do not satisfy superposition and shift invariance. But they should be causal and stable to be realized. Median filters and rank-order filters are very good examples for non-linear filters.
Comment: Very nice layout, read from tablet, Thanks Mr. Joshua
http:\\embedded-computing.com/
